A hierarchical collaborative optimization and settlement method for large-scale industrial demand response

Abstract

The application provides a hierarchical collaborative optimization and settlement method for large-scale industrial demand response, relates to the technical field of power systems, and comprises the following steps: dividing large-scale industrial demand response resources into multiple levels and constructing an optimization model; decomposing a global optimization problem into local sub-problems, defining consistent variable constraints for global consistency, and generating scheduling instructions through iterative calculation; obtaining actual response data in response to the scheduling instructions, performing hash operation and digital signature processing on the actual response data to generate a signed digest; verifying the signed digest by using a smart contract, executing real-time settlement after verification to obtain real-time settlement data, and summarizing to obtain periodic settlement data; calculating performance indicators according to the scheduling instructions and the actual response data, and performing numerical correction on the real-time settlement data and the periodic settlement data by using the performance indicators to obtain adjusted settlement data. The application realizes fast scheduling and convergence of large-scale resources, and effectively improves the enthusiasm and fairness of users participating in scheduling.

Description

Technical Field

[0001] This application relates to the field of power system technology, and in particular to a hierarchical collaborative optimization and settlement method for large-scale industrial demand response. Background Technology

[0002] Demand response is a crucial means of load-side resource regulation. It allows users to participate in system dispatch by adjusting their loads during periods of power supply and demand tension, thereby enhancing the flexibility and security of the power grid. In the area of ​​renewable energy generation, the potential for flexible adjustment of large-scale industrial loads has not yet been fully utilized. Due to the strong intermittent and fluctuating nature of renewable energy sources such as wind and solar power, the increasing scale of industrial loads presents new challenges to the power grid in balancing power and stabilizing voltage.

[0003] Current demand response management models primarily employ centralized scheduling and simple capacity-based settlement. Centralized scheduling involves grid operators or third-party aggregators centrally collecting data, optimizing it uniformly, and issuing scheduling instructions. This model relies on a central server to aggregate data from all terminals and solve the global optimization model, with communication links and computing nodes bearing the entire data interaction load. Regarding settlement, traditional methods typically involve manual reconciliation or periodic settlement based solely on declared capacity. Data transmission and recording lack unified encryption or verification mechanisms, and settlement data directly depends on the raw values ​​reported by the equipment.

[0004] Therefore, existing technologies are insufficient to simultaneously meet the requirements of rapid response to large-scale industrial resources and ensuring the authenticity and reliability of settlement data in real-time scheduling scenarios. Summary of the Invention

[0005] This application provides a hierarchical collaborative optimization and settlement method for large-scale industrial demand response to solve the problems mentioned in the background art.

[0006] The first aspect of this application provides a hierarchical collaborative optimization and settlement method for large-scale industrial demand response, including: dividing large-scale industrial demand response resources into multiple levels and constructing optimization models corresponding to the multiple levels;

[0007] The global optimization problem composed of the optimization model is decomposed into multiple local sub-problems. Consistency variables are defined to constrain the global consistency of multiple local sub-problems. Scheduling instructions are generated through iterative calculations between multiple levels.

[0008] In response to the scheduling instruction, the actual response data of the industrial demand response resources is obtained, and hash operation and digital signature processing are performed on the actual response data to generate a signature digest;

[0009] The signature digest is verified using a smart contract. After verification, real-time settlement is performed based on the actual response data to obtain real-time settlement data. The real-time settlement data within the target period is then aggregated and periodic settlement is performed to obtain periodic settlement data.

[0010] Define performance metrics, calculate the performance metrics based on scheduling instructions and actual response data, and perform numerical corrections on the real-time settlement data and the periodic settlement data based on the performance metrics to obtain adjusted settlement data.

[0011] Optionally, in one possible implementation of the first aspect, the plurality of layers includes an equipment layer, a plant layer, a zone layer, and a network-wide layer;

[0012] The device layer utilizes device agents to perform local control and status feedback of the target adjustable resources;

[0013] The plant layer uses a plant agent to aggregate information from the equipment layer and control the power balance within the plant.

[0014] The partitioning layer utilizes a partitioning agent to coordinate multiple plant / station layers within the same partition and generate partition-level targets.

[0015] The entire network layer performs network-wide security constraint management and cross-regional coordination.

[0016] Optionally, in one possible implementation of the first aspect, constructing an optimization model corresponding to multiple levels includes:

[0017] For the device layer, a physical constraint model is established to define the charging / discharging power boundary or the load power boundary;

[0018] For the plant layer, a plant power balance model is established to constrain the consistency between the total power of equipment within the plant and the external exchange power, and a plant operation efficiency model is established to optimize the operating cost of the plant.

[0019] For the partitioned layer, a partitioned power balance model is established to constrain the consistency between the sum of the switching power of the plants and stations within the partition and the partitioned interaction power, and a regional target model is established to smooth the power fluctuations within the partitioned layer.

[0020] For the entire network layer, a network power balance model is established to constrain the consistency between the sum of the switching power of each partition within the network and the total network interaction power, and a network operation efficiency model is established to optimize the network's operating costs and risks.

[0021] Optionally, in one possible implementation of the first aspect, the optimization model includes a local objective function and local constraints;

[0022] The process of generating scheduling instructions through iterative calculations across multiple levels includes:

[0023] The lower-level entity performs a solution based on the local objective function and the consistency variables generated in the previous iteration, obtains local decision variables, and uploads them to the upper-level entity.

[0024] The upper-level entity collects the local decision variables, updates the consistency variables based on global constraints, and distributes them to the lower-level entity.

[0025] The lower-level entity updates the Lagrange parameters corresponding to the consistency variables based on the updated consistency variables;

[0026] Repeat the steps of obtaining local decision variables, updating consistency variables, and updating Lagrange parameters until the preset convergence condition is met, at which point the scheduling instruction is output.

[0027] Optionally, in one possible implementation of the first aspect, generating scheduling instructions through iterative calculations across multiple levels further includes:

[0028] Obtain the preset prediction time domain length for the optimization model, and predict the load data and renewable energy output data within the prediction time domain length to obtain the prediction data;

[0029] Based on the predicted data, iterative calculations are performed to generate control sequences for multiple future time periods;

[0030] Extract the instruction corresponding to the current moment from the control sequence as the scheduling instruction;

[0031] The method further includes:

[0032] The scheduling instruction is issued, and the prediction data is updated at the next moment. The iterative calculation is then repeated based on the updated prediction data.

[0033] Optionally, in one possible implementation of the first aspect, performing hash operations and digital signature processing on the actual response data to generate a signature digest includes:

[0034] Perform a hash operation on the actual response data to generate digest data;

[0035] A digital signature is generated on the digest data using a private key;

[0036] The digest data and the digital signature are combined to form the signature digest.

[0037] Optionally, in one possible implementation of the first aspect, the step of using a smart contract to verify the signature digest, and then performing real-time settlement based on the actual response data to obtain real-time settlement data after successful verification, includes:

[0038] Parse the signature digest to verify the authenticity and integrity of the actual response data; when the verification is successful, calculate the power deviation between the actual response data and the scheduling instruction within a preset response window;

[0039] When the power deviation value is less than the preset tolerance threshold, the real-time conversion coefficient at the current moment is obtained, and the basic interaction value is calculated based on the real-time conversion coefficient at the current moment.

[0040] The real-time settlement data is generated by performing digital voucher allocation or data record update operations based on the basic interaction values ​​through smart contracts.

[0041] Optionally, in one possible implementation of the first aspect, the periodic settlement of the real-time settlement data within the aggregated target period is performed to obtain periodic settlement data, including:

[0042] All real-time settlement data generated within the target period are aggregated to obtain aggregated real-time settlement data.

[0043] Obtain the preset capacity baseline coefficient and promised response volume through smart contracts.

[0044] Calculate the capacity interaction value based on the preset capacity benchmark coefficient;

[0045] The performance interaction value is calculated based on the comparison between the cumulative actual response volume and the committed response volume within the target period.

[0046] The aggregated real-time settlement data, the capacity interaction value, and the performance interaction value are cumulatively calculated to generate the periodic settlement data.

[0047] Optionally, in one possible implementation of the first aspect, the performance metrics include response delay metrics, response consistency metrics, and output smoothness metrics.

[0048] The response delay index represents the time span from the moment the scheduling instruction is issued to the moment when the actual response data reaches the target response value specified by the scheduling instruction.

[0049] The response matching index characterizes the degree of overlap between the response curve corresponding to the actual response data within the preset response window and the target response curve corresponding to the scheduling instruction.

[0050] The output smoothness index characterizes the degree of dispersion of the actual response data during the response process.

[0051] Optionally, in one possible implementation of the first aspect, the step of performing numerical correction on the real-time settlement data and the periodic settlement data according to performance indicators to obtain adjusted settlement data includes:

[0052] The response delay index, response consistency index, and output smoothness index are all subjected to dimensionless processing.

[0053] Based on preset proportional parameters, the response time delay index, response consistency index and output smoothness index after dimensionless processing are combined and calculated to obtain comprehensive performance parameters.

[0054] The real-time settlement data and the periodic settlement data are numerically corrected using the comprehensive performance parameters to obtain the adjusted settlement data.

[0055] This application provides a hierarchical collaborative optimization and settlement method for large-scale industrial demand response, with the following beneficial effects:

[0056] 1. This application constructs an optimization model corresponding to multiple levels, clearly defining the responsibilities of each level, and defines consistency variables to constrain the global consistency of local subproblems, thus laying the foundation for optimization collaboration. Addressing the high computational complexity of large-scale systems, this application generates scheduling instructions through iterative calculations across multiple levels, effectively improving convergence efficiency and ensuring that scheduling results can quickly adapt to the dynamic changes of large-scale industrial resources.

[0057] 2. This application addresses the lack of reliable data verification in traditional settlement processes by generating a signature digest of the actual response data before settlement and using smart contracts for rigorous verification, ensuring that the data source involved in the settlement is authentic and has not been tampered with. Therefore, this application establishes a trust boundary for data interaction by adopting a settlement method based on cryptographic evidence, realizing closed-loop management of the entire process from instruction issuance and data feedback to real-time settlement, thereby improving the security and transparency of settlement data.

[0058] 3. This application uses indicators such as response delay, consistency, and output smoothness to make precise numerical corrections to real-time and periodic settlement data. This not only quantifies and distinguishes the adjustment quality of different resources, but also guides resource providers to proactively improve response speed and accuracy. Thus, while ensuring fairness, it constructs an efficient, scalable industrial demand response system with self-optimization capabilities. Attached Figure Description

[0059] Figure 1 This is a flowchart illustrating a hierarchical collaborative optimization and settlement method for large-scale industrial demand response provided in an embodiment of this application.

[0060] Figure 2 This is an overall logical function diagram of a hierarchical collaborative optimization and settlement method for large-scale industrial demand response provided in the embodiments of this application;

[0061] Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0062] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0063] The technical solutions of this application will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0064] To overcome the problems of slow response and lack of credibility and transparency in the results of large-scale industrial load scheduling in existing technologies, this application provides a hierarchical collaborative optimization and settlement method for large-scale industrial demand response. By generating scheduling instructions through hierarchical collaborative optimization and combining signature digest verification and performance index numerical correction strategies, real-time and periodic dual-track settlement is achieved under the premise of preventing data tampering during periods of rapid growth, thereby effectively improving the response speed and settlement credibility of large-scale industrial scheduling.

[0065] See Figure 1 This is a flowchart illustrating a hierarchical collaborative optimization and settlement method for large-scale industrial demand response provided in an embodiment of this application. Figure 1 The execution entity of the method shown can be a software and / or hardware device. The execution entity of this application can include, but is not limited to, at least one of the following: user equipment, network equipment, etc. User equipment can include, but is not limited to, computers, smartphones, personal digital assistants (PDAs), and the aforementioned electronic devices. Network equipment can include, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of computers or network servers. Cloud computing is a type of distributed computing, consisting of a super virtual computer composed of a group of loosely coupled computers. This embodiment does not limit this. Steps 100 to 400 are detailed below:

[0066] Step 100: Divide large-scale industrial demand response resources into multiple levels and construct optimization models corresponding to multiple levels.

[0067] Specifically, a hierarchical optimization framework is constructed, dividing large-scale industrial demand response resources into equipment layer, plant layer, zone layer, and network layer, realizing hierarchical scheduling management from local to global, and providing a clear organizational structure and computing boundaries for subsequent collaborative optimization.

[0068] It should be noted that the hierarchical optimization framework includes multiple levels, specifically the equipment level, plant level, zone level, and network-wide level.

[0069] The device layer utilizes device agents to perform local control and status feedback for the target adjustable resources. For example, each adjustable resource (load / energy storage / generation) is equipped with a device agent, which is mainly responsible for local execution, status feedback, and local constraints, such as power upper and lower limits, ramp rate, and minimum duration.

[0070] The plant layer uses the plant agent to aggregate information from the equipment layer and control the power balance within the plant. Specifically, it aggregates equipment information within the plant, performs local optimization and summarization, and is responsible for submitting aggregation capability and deviation information to the partition.

[0071] The partitioning layer uses a partitioning agent to coordinate multiple plant layers within the same partition and generate partition-level targets.

[0072] The network-wide layer performs network-wide security constraint management and cross-regional coordination, specifically responsible for network-wide security constraints, cross-regional coordination, emergency triggering, and global economic optimization, such as overall peak smoothing.

[0073] In some embodiments, step 100, "constructing an optimization model corresponding to multiple levels," specifically includes steps 110 to 140:

[0074] Step 110: For the device layer, establish a physical constraint model that defines the charging / discharging power boundary or the load power boundary.

[0075] A unified mathematical modeling system is constructed for different levels, clarifying the decision variables, optimization objectives and constraints of each level, to ensure that the hierarchical optimization has logical consistency and solvability.

[0076] Specifically, optimized modeling is performed at the equipment level, achieving operational modeling of individual devices such as industrial loads, energy storage devices, and renewable energy generation units at the lowest level, ensuring that the controllability and responsiveness of each device are accurately reflected. Specifically, a physical constraint model is established for each device; for example, for energy storage devices, charging and discharging power constraints need to be considered.

[0077] ;

[0078] in, Indicates device i At any moment t The discharge power, Indicates charging power. and These represent the upper limits of discharge and charging power, respectively.

[0079] For load equipment, a demand response model is used, such as adjustable load. satisfy:

[0080] ;

[0081] in, and These represent the minimum and maximum power of the load, respectively.

[0082] Step 120: For the plant level, establish a plant power balance model that constrains the consistency between the total power of equipment within the plant and the external exchange power, and a plant operation efficiency model for optimizing the operating cost of the plant.

[0083] Specifically, optimization modeling is performed at the plant / station level, unifying the management of all equipment within the plant / station to achieve power balance and locally optimal scheduling. The total power balance constraint of the plant / station is expressed as:

[0084] ;

[0085] in, Indicates the load aggregation within the plant. This represents a set of power generation units. For the first The power of the load, For the first The power of each power generation unit This indicates the power exchanged between the power plant and the external power grid.

[0086] To achieve energy balance and economic optimization within power plants, economic objectives are introduced at the power plant level, such as minimizing the power plant's electricity purchase cost. ,in, express t Electricity price at any given moment.

[0087] Step 130: For the partitioned layer, establish a partitioned power balance model that constrains the consistency between the sum of the switching power of the plants and stations within the partition and the partitioned interactive power, and a regional target model for smoothing power fluctuations within the partitioned layer.

[0088] Optimization modeling of the zonal layer is performed to coordinate the operation of multiple power plants within the distribution network zoning area, achieving cross-power plant energy coordination and load transfer. First, zonal power balance constraints are introduced:

[0089] ;

[0090] in, Indicates the set of plants and stations within the zone. Indicates factory station Power exchanged with the power grid This indicates the exchange power between the zone and the upper-level power grid, and the power plants / stations. Represents the set of plants and stations within the partition. The first in Each factory station.

[0091] Establish regional optimization objectives to improve the stability of local power grids, such as peak shaving and valley filling. ;in, This indicates the average power level of the partition.

[0092] Step 140: For the entire network layer, establish a network power balance model that constrains the consistency between the sum of the switching power of each partition within the network and the total network interaction power, and a network operation efficiency model for optimizing the network's operating costs and risks.

[0093] It should be noted that optimized modeling of the entire network layer is performed to achieve coordinated scheduling and optimization between zones across the entire distribution network, ensuring the safety and economy of the entire network operation. Specifically, power balance constraints are set for the entire network:

[0094] ;

[0095] in, Z Represents the set of all partitions. This indicates the total power exchanged between the entire network and the upstream power grid. Indicates partition z Power exchanged with the upstream power grid.

[0096] Furthermore, a comprehensive network optimization goal is constructed by combining economic efficiency and security. ;in, This represents the total operating cost of the entire network. Indicates the risk indicators of power grid operation. and These are the weighting coefficients.

[0097] Step 200: Decompose the global optimization problem composed of the optimization model into multiple local subproblems, define consistency variables to constrain the global consistency of multiple local subproblems, and generate scheduling instructions through iterative calculations between multiple levels.

[0098] It should be noted that the optimization model includes local objective functions and local constraints. Specifically, the optimization models established in step 100 are standardized into a form solvable by a distributed algorithm. The local objective functions correspond to the individual optimization objectives of each level (e.g., plant / station, zone), such as minimizing operating costs at the plant level and minimizing power fluctuations at the zone level. The local constraints correspond to the physical limitations within each level, such as the charging / discharging power boundaries of equipment and the power balance equations within the plant / station.

[0099] Based on the hierarchical optimization framework established in step 100, a distributed optimization algorithm, such as the Alternating Direction Method of Multipliers (ADMM), is introduced. Through parallel computing and information interaction between multiple levels, the scheduling results are quickly converged, ensuring high efficiency and scalability even with resources on a scale of millions.

[0100] The overall network optimization problem is decomposed into local subproblems that can be solved in parallel, and consistency variables are defined to ensure global consistency of solutions to different subproblems. The overall goal of the power grid is to reduce costs and losses while ensuring grid operation safety; therefore, the global optimization problem can be formulated as follows:

[0101] ;

[0102] in, Indicates device Decision variables (such as output and load adjustment). This represents a set of devices. This represents the adjustment cost function for the equipment. Indicates device At any moment The amount of power adjustment. This represents the network loss cost coefficient. Indicates time The power grid loss function. This represents the optimization time window. Constraints in the global optimization problem include network flow constraints. and node voltage constraints .in, This indicates the minimum voltage that a node is allowed to pass through. This indicates the minimum voltage that a node is allowed to pass through. Indicates time node The actual voltage.

[0103] Consistency must be maintained between different levels of the power grid. For example, the total output of power plants must equal the sum of the outputs of all equipment, and power exchange between zones must satisfy power flow constraints. To illustrate this, consistency variables are introduced. This transforms the global problem into a local problem and consistency constraints:

[0104] ;

[0105] in, Indicates factory station The set of local decision variables. This represents the local objective function of the plant. It is a mapping matrix, representing Map the plant's decision variables to a globally consistent quantity. It is a consistency variable, representing the aggregated indicators of the plant / station in the partition or the entire network. It is an indicator function used to ensure that consistency constraints are met.

[0106] Furthermore, this application utilizes the ADMM algorithm to iteratively converge each level to a globally feasible solution, and supports parallel and asynchronous computation. The iterative steps of ADMM are as follows (the number of iterations is denoted as ). k ).

[0107] In some embodiments, step 200, "generating scheduling instructions through iterative calculations across multiple levels," specifically includes steps A1 to A4:

[0108] Step A1: The lower-level entity performs a solution based on the local objective function and the consistency variables generated in the previous iteration, obtains local decision variables, and uploads them to the upper-level entity.

[0109] Lower-level updates (equipment / plant level): Each plant independently resolves its own issues.

[0110] ;

[0111] in, Indicates factory station The local objective function. It is a penalty parameter used to balance consistency constraints. It is the first k The Lagrange multipliers in the next iteration represent the penalty for the consistency constraint. It is the first Consistency variables during the iteration process.

[0112] Step A2: The upper-level entity collects the local decision variables, updates the consistency variables based on global constraints, and distributes them to the lower-level entity.

[0113] Upper-layer update (partition / network-wide): After collecting results from all plants, the partition / network-wide update is responsible for adjusting consistency variables. :

[0114] ;

[0115] in, This represents a global constraint function, such as voltage and power flow safety constraints.

[0116] Step A3: The lower-level entity updates the Lagrange parameters corresponding to the consistency variables based on the updated consistency variables.

[0117] During each iteration, each plant needs to update the Lagrange multipliers, and the update process is as follows:

[0118] ;

[0119] Each plant only needs to use local data, equipment characteristics, and local cost functions, meaning that plant calculations are independent of each other, thus enabling high parallelism. If some plants experience delays, asynchronous reporting is also allowed, and the partition layer continues calculations using existing results, avoiding blocking.

[0120] Step A4: Repeat the steps of obtaining local decision variables, updating consistency variables, and updating Lagrange parameters, that is, repeat steps A1 to A3 until the preset convergence condition is met and the scheduling instruction is output.

[0121] The preset convergence condition refers to the condition during algorithm iteration where the residual norm of the consistency variable between two consecutive iterations is less than a preset convergence threshold, such as 10. -4 When the number of iterations reaches the preset maximum number of iterations, or when any of the above conditions are met, the algorithm is considered to have converged and the final scheduling instruction is output.

[0122] To improve the robustness of the algorithm in real-world environments, enabling it to converge efficiently even with communication delays, data loss, or device offline conditions, the matrix... and factory station Local objective function Normalization is performed to avoid numerical instability and reduce iterative oscillations. Simultaneously, each iteration uses the previous optimization result as the initial value for the current iteration, avoiding starting from zero and significantly accelerating convergence. To ensure overall stability, if a device goes offline, the predicted value can be used instead, or compensation can be provided by redundant resources in the partition.

[0123] In some embodiments, generating scheduling instructions through iterative calculations across multiple levels may further include steps B1 to B3:

[0124] Step B1: Obtain the preset prediction time domain length for the optimization model, and predict the load data and renewable energy output data within the prediction time domain length to obtain the prediction data.

[0125] Step B2: Perform iterative calculations based on the predicted data to generate control sequences for multiple future time periods.

[0126] Step B3: Extract the instruction corresponding to the current moment from the control sequence as the scheduling instruction.

[0127] This application also includes issuing scheduling instructions to update the prediction data at the next time step, and repeatedly performing iterative calculations based on the updated prediction data.

[0128] It should be noted that in dynamic and uncertain power grid environments, such as load fluctuations, unstable renewable energy output, and changes in equipment status, rolling predictive optimization is introduced on the basis of the existing distributed optimization framework to ensure that the optimized scheme always remains consistent with the actual situation, thereby improving real-time performance and robustness. The MPC (Model Predictive Control) method is adopted, and rolling predictive optimization is performed every [period]. Run a distributed optimization once per time interval. Set the prediction time domain length to... Regarding the future , , ..., Forecast load and renewable energy output:

[0129] ;

[0130] in, This represents the predicted load power. This represents the projected renewable energy power.

[0131] The time domain prediction is optimized using the framework outlined above, with constraints including power flow constraints, voltage constraints, and consistency constraints.

[0132] ;

[0133] The optimization result is a series of control sequences, such as power adjustments every 5 minutes. Only the current time-to-time sequence is actually executed. Control command. Proceed to the next moment. Then, based on the latest data, such as actual measurements and new forecasts, the optimization is repeated. This can correct forecast errors and avoid deviations caused by inaccurate long-term forecasts.

[0134] Step 300: In response to the scheduling instruction, obtain the actual response data of the industrial demand response resources, and perform hash operation and digital signature processing on the actual response data to generate a signature digest.

[0135] Combining the optimization results obtained in step 200, the optimization results are combined with the execution data using the blockchain smart contract mechanism. A trusted settlement boundary is established between different levels, and the dual-track operation of real-time settlement and periodic settlement is automatically completed, realizing the transparency and automation of the settlement process.

[0136] Define the main entities of the smart contract, including equipment agents, plant agents, aggregators, regional agents, distribution network operators, and third-party metering / auditing institutions. The content elements of the contract are represented using structured fields, as shown in Table 1.

[0137] Table 1

[0138]

[0139] For privacy reasons, only the digest / hash and key statistics (signed) are written on-chain. Raw measurement data is stored off-chain and audited as needed. The minimum on-chain information is sufficient to verify the correctness of the aggregation without disclosing the original sensitive data, while off-chain retention also meets the requirements for measurement and verification. On-chain information is public, and sensitive off-chain data is only provided to the reviewer designated by the contract.

[0140] It should be noted that step 300, "performing hash operations and digital signature processing on the actual response data to generate a signature digest," specifically includes steps C1 to C3.

[0141] Step C1: Perform a hash operation on the actual response data to generate digest data.

[0142] Step C2: Generate a digital signature for the digest data using the private key.

[0143] Step C3: Combine the digest data and digital signature into the signature digest, and write the signature digest into the distributed ledger.

[0144] It should be noted that this application enables efficient interaction of data and control commands between different levels (equipment layer, plant layer, zone layer, and network layer) through data and control interfaces, ensuring that the optimization framework can operate smoothly under the collaboration of multiple entities.

[0145] Define a standardized Application Programming Interface (API), including functions such as data requests, status updates, and optimization command issuance. The interface must conform to the RESTful API structure to ensure cross-platform compatibility and scalability.

[0146] JSON (JavaScript Object Notation) or Protobuf (Protocol Buffers) is used as the data exchange format. For example, the JSON format for information reported from the plant layer to the partition layer is as follows:

[0147] {

[0148] "station_id": "S12",

[0149] "timestamp": "2025-09-14T08:00:00Z",

[0150] "net_power": 3.5,

[0151] "reserve_capacity": 1.2

[0152] }

[0153] Lightweight publish / subscribe communication is achieved using the MQTT (Message Queuing Telemetry Transport) protocol, supporting concurrent access from a large number of devices.

[0154] Furthermore, this application uses a hash encryption mechanism to ensure data traceability and tamper-proofness, achieving the aggregation and mapping of data and optimization results from the device layer to the entire network layer. A mapping function is defined. Aggregate equipment-level data into plant-level metrics:

[0155] ;

[0156] in, It can be a linear summation or a weighted aggregation function. This refers to the fact that the factory or station belongs to the factory or station. This is a collection of all adjustable resources under its jurisdiction (such as load, energy storage, etc.). Similarly, plant-level indicators are aggregated into regional indicators to form network-wide indicators.

[0157] To ensure the security, transparency, and tamper-proof nature of data exchange and optimization results, and to provide a trusted data foundation for subsequent smart contract settlements, a hash value is generated for each optimization result, for example, using SHA-256 (Secure Hash Algorithm 256-bit):

[0158] ;

[0159] in, M This represents the set of input and output data for a certain optimization. This corresponds to a hash value. The hash value is stored in a distributed ledger to ensure the data is immutable.

[0160] Furthermore, this application uses a proof chain to ensure that on-chain triggered settlements are based on authentic and immutable data, compensating for the differences in credibility between off-chain device data, and proving that a record indeed originated from a specific device or point in time in case of disputes. To ensure that a device cannot deny its submission at a certain moment, signatures are used to guarantee non-repudiation. Each device generates a small block of raw data with each response. Calculate the hash value accordingly. :

[0161] ;

[0162] in, This represents a hash function.

[0163] Specifically, the device utilizes the elliptic curve digital signature algorithm. Sign the hash to obtain the final digital signature. This is used to prove to smart contracts and higher-level entities that the data indeed originated from the entity and has not been tampered with.

[0164] ;

[0165] in, Indicates device i At any moment t The final digital signature generated.

[0166] The contract executes settlement after verifying the digest and signature via an Oracle (trusted data source) or multi-signature verification. The digest and signature are recorded on the blockchain, and the original data is only decrypted and provided according to permissions during arbitration or auditing.

[0167] Step 400: Verify the signature digest using a smart contract. After successful verification, perform real-time settlement based on the actual response data to obtain real-time settlement data, and summarize the real-time settlement data within the target period to perform periodic settlement to obtain periodic settlement data.

[0168] It should be noted that the real-time scheduling (such as power adjustment) given in the aforementioned steps... To ensure rapid response and timely reward disbursement, participants are encouraged to respond proactively to emergencies. For speed, small payments are often made quickly based on the signature summary submitted by the factory / station, i.e., real-time settlement.

[0169] In some embodiments, step 400 is specifically implemented by steps 410 and 420:

[0170] Step 410: Verify the signature digest using a smart contract. Once verified, perform real-time settlement based on the actual response data to obtain real-time settlement data.

[0171] Specifically, step 410 includes steps 411 to 414:

[0172] Step 411: Parse the signature digest to verify the authenticity of the source and integrity of the actual response data.

[0173] Step 412: In response to successful verification, calculate the power deviation between the actual response data and the scheduling command within the preset response window.

[0174] Step 413: In response to the power deviation value being less than the preset tolerance threshold, obtain the real-time conversion coefficient at the current moment, and calculate the basic interaction value based on the real-time conversion coefficient at the current moment.

[0175] Step 414: Execute digital voucher allocation or data record update operations based on the basic interaction values ​​through smart contracts to generate real-time settlement data.

[0176] It should be noted that when the partition layer (managed by the partition agent) issues a fast adjustment command, a response time window is set, denoted as . At the end of the window, check the actual power change for each device. ,equipment At any moment The variation in load or power generation output. A preset tolerance threshold is provided. This indicates that minor deviations will not be included in the settlement, preventing unfair treatment of equipment due to measurement errors or performance fluctuations. If the equipment... Actual response If it meets the triggering conditions for real-time settlement, then it is considered to have met the conditions.

[0177] At any moment ,equipment Real-time settlement amount Let it be denoted as , and the calculation formula is as follows:

[0178] ;

[0179] in, Indicates partitioning layer by time This is the real-time settlement unit price for the initial settlement period (e.g., [t, t+Δt)). The real-time settlement unit price is generated synchronously by the partition layer during the hierarchical collaborative optimization iteration in step 200, reflecting the marginal power regulation value of that period. This indicates that if the response amount is less than the threshold If the response volume exceeds the threshold, it will not be counted as a reward or trigger; if the response volume exceeds the threshold, it will be deducted. This is calculated later. This method ensures that the rewards received by the device are positively correlated with its effective response volume and avoids settlement anomalies caused by small fluctuations.

[0180] The execution process of a smart contract is as follows: the device layer or plant layer agent transmits its execution result. The data is submitted to the partitioning layer proxy via a signature and hash digest. The partitioning layer proxy aggregates the data and generates a signed execution proof (containing both a hash digest and a signature). The partitioning layer proxy submits the execution proof to the smart contract. After the contract verifies the data's legitimacy (the digest and signature match), it automatically executes the transfer or accounting operation, completing real-time settlement. This ensures real-time settlement without human intervention, avoids disputes, and improves efficiency. Simultaneously, the hash and signature mechanism guarantees data authenticity and immutability.

[0181] Step 420: Summarize the real-time settlement data within the target period and perform periodic settlement to obtain periodic settlement data.

[0182] Specifically, step 420 includes steps 421 to 425:

[0183] Step 421: Aggregate all real-time settlement data generated within the target period to obtain aggregated real-time settlement data.

[0184] Step 422: Obtain the preset capacity baseline coefficient and promised response amount through smart contracts.

[0185] Step 423: Calculate the capacity interaction value based on the preset capacity benchmark coefficient.

[0186] Step 424: Calculate the performance interaction value based on the comparison result of the cumulative actual response volume and the promised response volume within the target period.

[0187] Step 425: Perform cumulative calculation on the aggregated real-time settlement data, the capacity interaction value, and the performance interaction value to generate the periodic settlement data.

[0188] Specifically, the "prepayment / temporary settlement" in real-time settlement is aggregated, reconciled, and periodically settled with all transactions within the period. Periodic settlement is the final clearing, based on more rigorous measurement and verification. Firstly, during the collection period... Signature, already occurred Records (on-chain / off-chain) and aggregate variables Next, based on the capacity agreed upon in the current date and the contract. Pay the basic capacity fee, i.e., capacity settlement:

[0189] ;

[0190] in, This indicates the contract price or market clearing price.

[0191] Then based on the actual execution volume within the period Performance is settled by comparing with commitments, using either a stepped or continuous function. Payment / Deduction :

[0192] ;

[0193] The total settlement amount can be expressed as:

[0194] ;

[0195] in, This indicates the total real-time settlement amount. This indicates that the fees may include aggregator fees, distribution network service fees, taxes, etc. This indicates a fine.

[0196] When a device or user objects to the settlement result, the smart contract triggers an arbitration process and writes the disputed event to the blockchain. A third-party measurement / auditing agency, according to its permissions, retrieves the original off-chain data and provides a verification report. The arbitration result is executed through the contract, such as adjusting payments, recording transactions, or imposing penalties. After review and confirmation that there is no dispute, the contract triggers the actual transfer and writes the final invoice summary onto the blockchain.

[0197] Step 500: Define performance indicators, calculate performance indicators based on scheduling instructions and actual response data, and perform numerical corrections on real-time settlement data and periodic settlement data based on performance indicators to obtain adjusted settlement data.

[0198] Specifically, performance metrics include response time delay, response consistency, and output smoothness.

[0199] The response delay index represents the time span from the moment the scheduling instruction is issued to the moment when the actual response data reaches the target response value specified by the scheduling instruction.

[0200] The response matching index characterizes the degree of overlap between the response curve corresponding to the actual response data within the preset response window and the target response curve corresponding to the scheduling instruction.

[0201] The output smoothness index characterizes the degree of dispersion of the actual response data during the response process.

[0202] For example, to accurately quantify the response quality of each participant and make it usable for calculating settlement coefficients, performance metrics are defined, including response latency. Response accuracy and stability Response delay This indicates the time from the issuance of the scheduling command to the device reaching the specified response value (e.g., reaching 90% of the target value). Response accuracy. This represents the mean square error or relative error between the actual response and the target response within a specified window. For example:

[0203] ;

[0204] in, Indicates response to the assessment window, This is the window length. Indicates device At any moment The actual power adjustment amount. Indicates device At any moment The target power adjustment amount.

[0205] stability Indicators representing response volatility are used to measure the smoothness and stability of the response process, such as the standard deviation of actual power adjustment. .

[0206] In some embodiments, step 500, "performing numerical corrections on real-time settlement data and periodic settlement data based on performance indicators to obtain adjusted settlement data," specifically includes steps D1 to D3:

[0207] Step D1: Perform dimensionless processing on the response delay index, response consistency index, and output smoothness index respectively.

[0208] Step D2: Based on the preset proportional parameters, perform combined calculations on the dimensionless processing of the response delay index, response consistency index, and output smoothness index to obtain the comprehensive performance parameters.

[0209] Step D3: Perform numerical correction on the real-time settlement data and the periodic settlement data using the comprehensive performance parameters to obtain the adjusted settlement data.

[0210] Specifically, performance metrics are mapped to settlement performance factors. This allows high-quality responses to receive higher pay, while low-quality responses are discounted. Firstly, performance metrics... , , Standardize to [0,1], then represent the performance factor using a weighted combination:

[0211] ;

[0212] in, , , Represents the weighting coefficient, and . , , This represents the sensitivity parameter.

[0213] The adjusted final payment is obtained using performance factors:

[0214] ;

[0215] in, It can be linear amplification or tiered reward. This indicates the total real-time settlement. This refers to capacity settlement within periodic settlements. Performance settlement in periodic settlement This indicates that the fees may include aggregator fees, distribution network service fees, taxes, etc. This indicates a fine.

[0216] Furthermore, to ensure the time consistency and verifiability of delay and accuracy measurements, the maximum permissible time deviation is... Performance metric accuracy needs to be met (e.g.) The scheduling instructions also include a unique ID and a timestamp. The device execution message includes the execution time. With signature, response delay Need to meet For precision measurements, cleaned [materials / materials] are used. With the target specified in the contract .

[0217] Furthermore, to ensure the performance evaluation mechanism is executable and reduces false positives or abuse, a fault tolerance threshold can be set, such as latency tolerance. and precision threshold Exceeding the threshold will trigger downgrade or arbitration. In addition, a "temporary reporting period" is provided before settlement for user review; for example, disputes can be raised within 1 hour to trigger arbitration in real-time settlement.

[0218] Based on the above steps, this application also includes the following embodiments:

[0219] See Figure 3 This is a schematic diagram of the hardware structure of an electronic device provided in an embodiment of this application. The electronic device 40 includes: a processor 41, a memory 42, and a computer program.

[0220] The memory 42 is used to store computer programs. The memory can also be flash memory. The computer programs are, for example, applications or functional modules that implement the above methods.

[0221] The processor 41 is used to execute the computer program stored in the memory to implement the various steps performed by the device in the above method, as can be seen from the relevant descriptions in the preceding method embodiments.

[0222] Alternatively, the memory 42 can be either standalone or integrated with the processor 41.

[0223] When the memory 42 is a device independent of the processor 41, the device may also include:

[0224] Bus 43 is used to connect the memory 42 and the processor 41.

[0225] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A hierarchical collaborative optimization and settlement method for large-scale industrial demand response, characterized in that, include: Large-scale industrial demand response resources are divided into multiple levels, and optimization models corresponding to these multiple levels are constructed. The multiple levels include the equipment layer, the plant / station layer, the zone layer, and the entire network layer; The device layer utilizes device agents to perform local control and status feedback of the target adjustable resources; The plant layer uses a plant agent to aggregate information from the equipment layer and control the power balance within the plant. The partitioning layer utilizes a partitioning agent to coordinate multiple plant / station layers within the same partition and generate partition-level targets. The entire network layer performs network-wide security constraint management and cross-regional coordination. The optimization model includes a local objective function and local constraints. The global optimization problem, composed of the aforementioned optimization model, is decomposed into multiple local subproblems. Consistency variables are defined to constrain the global consistency of these local subproblems. Scheduling instructions are generated through iterative calculations across multiple levels, including: The lower-level entity performs a solution based on the local objective function and the consistency variables generated in the previous iteration, obtains local decision variables, and uploads them to the upper-level entity. The upper-level entity collects the local decision variables, updates the consistency variables based on global constraints, and distributes them to the lower-level entity. The lower-level entity updates the Lagrange parameters corresponding to the consistency variables based on the updated consistency variables; Repeat the steps of obtaining local decision variables, updating consistency variables, and updating Lagrange parameters until the preset convergence condition is met, and then output the scheduling instruction. In response to the scheduling instruction, the actual response data of the industrial demand response resources is obtained, and hash operation and digital signature processing are performed on the actual response data to generate a signature digest; The signature digest is verified using a smart contract. After verification, real-time settlement is performed based on the actual response data to obtain real-time settlement data. The real-time settlement data within the target period is then aggregated and periodic settlement is performed to obtain periodic settlement data. Define performance metrics and calculate the performance metrics based on scheduling instructions and actual response data. The performance metrics include response delay metrics, response consistency metrics, and output smoothness metrics. The response delay index represents the time span from the moment the scheduling instruction is issued to the moment when the actual response data reaches the target response value specified by the scheduling instruction. The response matching index characterizes the degree of overlap between the response curve corresponding to the actual response data within the preset response window and the target response curve corresponding to the scheduling instruction. The output smoothness index characterizes the degree of dispersion of the actual response data during the response process; Adjusted settlement data is obtained by performing numerical corrections on the real-time settlement data and the periodic settlement data based on performance indicators, including: The response delay index, response consistency index, and output smoothness index are all subjected to dimensionless processing. Based on preset proportional parameters, the response time delay index, response consistency index and output smoothness index after dimensionless processing are combined and calculated to obtain comprehensive performance parameters. The real-time settlement data and the periodic settlement data are numerically corrected using the comprehensive performance parameters to obtain the adjusted settlement data.

2. The method according to claim 1, characterized in that, The construction of an optimization model corresponding to multiple levels includes: For the device layer, a physical constraint model is established to define the charging / discharging power boundary or the load power boundary; For the plant layer, a plant power balance model is established to constrain the consistency between the total power of equipment within the plant and the external exchange power, and a plant operation efficiency model is established to optimize the operating cost of the plant. For the partitioned layer, a partitioned power balance model is established to constrain the consistency between the sum of the switching power of the plants and stations within the partition and the partitioned interaction power, and a regional target model is established to smooth the power fluctuations within the partitioned layer. For the entire network layer, a network power balance model is established to constrain the consistency between the sum of the switching power of each partition within the network and the total network interaction power, and a network operation efficiency model is established to optimize the network's operating costs and risks.

3. The method according to claim 1, characterized in that, The method further includes: The scheduling instructions are generated through iterative calculations across multiple levels, and also include: Obtain the preset prediction time domain length for the optimization model, and predict the load data and renewable energy output data within the prediction time domain length to obtain the prediction data; Based on the predicted data, iterative calculations are performed to generate control sequences for multiple future time periods; Extract the instruction corresponding to the current moment from the control sequence as the scheduling instruction; The method further includes: The scheduling instruction is issued, and the prediction data is updated at the next moment. The iterative calculation is then repeated based on the updated prediction data.

4. The method according to claim 1, characterized in that, The step of performing hash operations and digital signature processing on the actual response data to generate a signature digest includes: Perform a hash operation on the actual response data to generate digest data; A digital signature is generated on the digest data using a private key; The digest data and the digital signature are combined to form the signature digest.

5. The method according to claim 1, characterized in that, The process of using smart contracts to verify the signature digest, and then performing real-time settlement based on the actual response data to obtain real-time settlement data, includes: Parse the signature digest to verify the authenticity and integrity of the actual response data; when the verification is successful, calculate the power deviation between the actual response data and the scheduling instruction within a preset response window; When the power deviation value is less than the preset tolerance threshold, the real-time conversion coefficient at the current moment is obtained, and the basic interaction value is calculated based on the real-time conversion coefficient at the current moment. The real-time settlement data is generated by performing digital voucher allocation or data record update operations based on the basic interaction values ​​through smart contracts.

6. The method according to claim 1, characterized in that, The periodic settlement data obtained by performing periodic settlement on the real-time settlement data within the target period includes: aggregating all real-time settlement data generated within the target period to obtain aggregated real-time settlement data; Obtain the preset capacity baseline coefficient and promised response volume through smart contracts; Calculate the capacity interaction value based on the preset capacity benchmark coefficient; The performance interaction value is calculated based on the comparison between the cumulative actual response volume and the committed response volume within the target period. The aggregated real-time settlement data, the capacity interaction value, and the performance interaction value are cumulatively calculated to generate the periodic settlement data.

Images