A method and system for remote control and safety verification of new energy equipment

Abstract

This invention relates to the field of IoT security control technology, solving the technical problems of existing technologies that are unaware of transient issues such as overshoot and oscillation, cannot automatically compensate for steady-state errors, and do not consider the dynamic changes in response delay. In particular, it relates to a method and system for remote control and security verification of new energy equipment. The method involves the following steps: acquiring the original control commands of the target device; processing the original control commands through parsing verification and initial permission screening to obtain parsed control commands; detecting overshoot through sampling sequences and triggering closed-loop correction; directly using the measured response delay for delay compensation, thus significantly shortening the response time; eliminating static errors and adapting to changes in communication delay; and significantly reducing the correction failure rate caused by delay jitter, thereby significantly improving control efficiency and robustness. It employs iterative convergence judgment and correction count recording to achieve a verifiable function with extremely high accuracy.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) security control technology, and in particular to a method and system for remote control and security verification of new energy equipment. Background Technology

[0002] IoT security control refers to the use of key technologies such as authentication mechanisms, key management, encrypted transmission, access control, secure routing protocols, intrusion detection, and fault tolerance within the IoT architecture. Existing technologies involve issuing start / stop or power adjustment commands through a human-machine interface, encapsulating and transmitting these commands via protocols, deploying industrial firewalls at the boundary in some systems for basic protocol filtering and blocking of malicious commands, performing simple format verification and permission checks, confirming the execution results of the commands, and finally storing control logs in a local database or relational database for later review.

[0003] Existing technologies lack millisecond-level continuous feedback sampling, making it impossible to acquire the dynamic deviation trajectory during the device's response process. This results in an inability to detect transient issues such as overshoot and oscillation, and an inability to automatically compensate for them. Furthermore, they fail to consider the dynamic changes in response delay and cannot distinguish between equipment failure and improper control parameters. For example, a certain energy storage PCS may experience a 5kW steady-state error between the actual power and the command due to temperature changes. In special cases where communication network delay fluctuates randomly, such as a sudden change from 20ms to 200ms, the open-loop control of existing technologies cannot eliminate steady-state errors caused by environmental changes, such as those caused by temperature drift. The lack of adaptive compensation capability for random fluctuations in communication network delay leads to over-adjustment of the correction amount, which can easily cause power oscillations. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and system for remote control and safety verification of new energy equipment. It solves the technical problems of existing technologies that are unaware of transient problems such as overshoot and oscillation, cannot automatically compensate for steady-state errors, do not consider the dynamic changes in response delay, and cannot accurately identify software and hardware faults.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for remote control and security verification of new energy equipment, the method comprising the following steps: collecting the original control commands of the target equipment, and obtaining the parsed control commands by parsing, verifying and initially screening permissions based on the original control commands; Based on the parsed control instructions, multi-source queue candidate processing is performed to obtain candidate arbitration instructions. Conflict detection arbitration processing is then performed based on the candidate arbitration instructions to obtain priority arbitration instructions. Health satisfaction analysis is performed on the priority arbitration instructions to obtain the equipment health feature vector and resource satisfaction. Based on the equipment health feature vector and resource satisfaction, the pre-verification pass instructions are obtained through the verification deviation process. Based on the pre-verification, high-precision delay processing is performed through instructions to obtain the instantaneous deviation vector and the measured response delay. Steady-state deviation correction accuracy processing is then performed on the instantaneous deviation vector and the measured response delay to obtain the execution feedback result. Based on the execution feedback results, audit logs are obtained through multi-source auditing. Among them, high-precision delay processing based on pre-verification through instructions includes: high-precision processing of instructions based on pre-verification to obtain instruction issuance records and initial device status; Based on the instructions issued, the records are continuously collected and processed to generate a real-time feedback sequence; The real-time feedback sequence and pre-verification are processed by control deviation delay through instructions to obtain the instantaneous deviation vector and the measured response delay.

[0006] Preferably, the process of parsing, verifying, and initially screening permissions based on the original control instructions includes: obtaining preliminary parsing instructions by parsing and format verification based on the original control instructions; Based on the initial parsing instructions, perform two-factor authentication and token verification to generate token authentication instructions; The token authentication command is initially screened for least privilege to obtain the parsed control command.

[0007] Preferably, the multi-source queue candidate processing based on the parsed control instructions includes: extracting a multi-source instruction set by taking a snapshot of the multi-source queue state based on the parsed control instructions; Dynamic priority weights are obtained by performing dynamic priority processing based on multi-source instruction sets. Arbitration output candidate processing is performed on the dynamic priority weight and multi-source instruction set to obtain candidate arbitration instructions.

[0008] Preferably, conflict arbitration processing is performed according to the candidate arbitration instruction, including: processing the conflict level according to the candidate arbitration instruction to obtain the conflict detection result; Based on the conflict detection results and candidate arbitration instructions, the resolution instructions are processed to obtain the resolved instructions; The resolved instructions are subjected to priority arbitration to obtain priority-arbitrated instructions.

[0009] Preferably, a health satisfaction analysis is performed on the priority arbitration instructions, including: processing the priority arbitration instructions with health feature parameters to obtain the equipment health feature vector and resource parameter vector; Based on the equipment health feature vector and priority arbitration instructions, the equipment health degradation amount is processed to obtain the predicted health degradation amount; Resource condition satisfaction is analyzed based on the resource parameter vector and priority arbitration instructions to obtain resource satisfaction.

[0010] Preferably, the deviation verification process based on the equipment health feature vector and resource satisfaction includes: generating a historical deviation sequence based on the deviation sequence of the equipment health feature vector and resource parameter vector; Adaptive threshold processing is performed based on historical deviation sequences to generate adaptive threshold factors. The adaptive threshold factor, predicted health degradation, and resource satisfaction are comprehensively verified to obtain the pre-verification pass instruction.

[0011] Preferably, the steady-state deviation correction accuracy processing of the instantaneous deviation vector and the measured response delay includes: obtaining the steady-state prediction deviation by processing the instantaneous deviation vector through steady-state deviation trend processing; Compensation and correction are performed based on steady-state prediction deviation and measured response time delay to obtain the compensation and correction amount; The compensation correction amount and the pre-verification are subjected to secondary correction and accuracy processing through instructions to obtain the execution feedback result.

[0012] Preferably, the multi-source audit process is performed based on the execution feedback results, including: generating a consistency comparison report by comparing the consistency of multi-source results based on the execution feedback results; Based on the consistency comparison report, perform digital non-repudiation signature processing to obtain the signed record; The signed records are subjected to blockchain auditing to obtain audit logs.

[0013] Preferably, the original control command includes the target device identifier, operation type, operation parameters, source priority level, and request timestamp.

[0014] This technical solution also provides a remote control and safety verification system for new energy equipment, which includes: The parsing module is used to collect the original control commands of the target device, and to obtain the parsed control commands by parsing, verifying and initially screening the permissions based on the original control commands. The priority arbitration module is used to process multi-source queue candidates based on parsed control instructions to obtain candidate arbitration instructions, and to perform conflict detection arbitration processing based on the candidate arbitration instructions to obtain priority arbitration instructions. The pre-verification module is used to perform health satisfaction analysis on the instructions after priority arbitration, obtain the equipment health feature vector and resource satisfaction, and obtain the pre-verification pass instruction based on the equipment health feature vector and resource satisfaction through the verification deviation verification process. The execution feedback module is used to perform high-precision delay processing based on the pre-verification through instructions to obtain the instantaneous deviation vector and the measured response delay. The instantaneous deviation vector and the measured response delay are then processed for steady-state deviation correction accuracy to obtain the execution feedback result. The audit module is used to obtain audit logs through multi-source audit processing based on the execution feedback results.

[0015] By employing the above technical solution, the present invention provides a method and system for remote control and safety verification of new energy equipment, which has at least the following beneficial effects: 1. This invention not only considers source priority levels but also introduces three dynamic factors: equipment health score, emergency flag, and instruction waiting time. This avoids indefinite instruction backlog and not only achieves dynamic priority arbitration but also links with subsequent equipment health prediction models. By identifying mutual exclusion and directional conflicts through instruction collision detection algorithms and calculating the comprehensive score of each instruction, the deviation tolerance can be adjusted based on the flag. This reduces the number of corrections without sacrificing security, thereby improving control efficiency. The introduction of a metadata encapsulation algorithm enables full traceability of the arbitration process. Arbitration metadata is directly used as input for digital signatures, enhancing the compliance and non-repudiation of the entire process.

[0016] 2. This invention employs a long short-term memory network, using core device temperature, vibration amplitude, cumulative start-stop count, and capacity decay rate as input features. Combined with operation type and operation parameters, it addresses the issue of wasted available resources and the misplacement of devices with high health scores but rapidly increasing degradation after execution due to local anomalies, leading to safety hazards. This invention not only integrates energy storage SOC and power limits but also links with environmental monitoring, significantly reducing the false rejection rate while maintaining the dangerous command interception rate. By comprehensively judging and comparing the predicted degradation with adaptive thresholds and resource satisfaction, it enhances the interpretability and fault tolerance of the entire control system.

[0017] 3. This invention detects overshoot through sampling sequences, triggers closed-loop correction, and uses the measured response delay directly for delay compensation, enabling the correction amount to predict lag in advance, significantly reducing overshoot and shortening response time. By dynamically adjusting the correction amount based on the difference between the measured response delay and the theoretical delay, it not only eliminates static errors but also adapts to changes in communication delay. Furthermore, it significantly reduces the correction failure rate caused by delay jitter, substantially improving control efficiency and robustness. It employs iterative convergence judgment and correction count recording to achieve verifiable functionality with extremely high accuracy, reducing fault location time from hours to minutes, thus integrating control and diagnosis.

[0018] 4. This invention combines the consistency comparison report with the execution feedback result to form a signed record containing a complete control chain. Any node can use the public key to verify the authenticity of the signature, further enhancing the mutual trust of cross-domain auditing. It can also solve the problems of easy log tampering and single point of failure, realizing full-link traceability from instruction initiation to final audit. In special scenarios such as hidden fault identification, cross-domain mutual trust, and batch verification, it demonstrates comprehensive advantages of high security, high compliance, and high efficiency. Attached Figure Description

[0019] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart of a remote control and safety verification method for new energy equipment according to the present invention; Figure 2 This is a structural block diagram of a remote control and safety verification system for new energy equipment according to the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0021] Example 1: Because current technology is unaware of transient issues such as overshoot and oscillation, cannot automatically compensate for steady-state errors, does not consider dynamic changes in response delay, and cannot accurately identify specific problems, please refer to [the relevant documentation / reference]. Figure 1 This embodiment provides a method for remote control and safety verification of new energy equipment, which can detect transient problems, compensate for steady-state errors, and accurately identify specific software and hardware problems. The method includes the following steps: S1. Collect the original control commands of the target device, and obtain the parsed control commands by parsing, verifying and initially screening permissions based on the original control commands. Existing technologies lack a strict decoding mechanism based on state machines, making them vulnerable to network attacks such as malformed packets and buffer overflows. They rely solely on static passwords or fixed tokens, which cannot effectively resist replay attacks and man-in-the-middle attacks. Furthermore, they lack dynamic binding with the command content and do not specify the combination of specific device types and operation types, resulting in a high risk of unauthorized operation. To solve the above problems, the specific implementation steps are as follows: S11. Based on the original control command, a preliminary parsed command is obtained through parsing and format verification. In this step, the state machine-based command decoding algorithm receives the original control command as input. By traversing the protocol state sequence of the command frame byte by byte, it first extracts the frame header and length field, and then reads the original encoded values ​​of the target device unique identifier, operation type, operation value, source priority level, and request timestamp in sequence. The algorithm performs a summation operation on all fields except the check code, and compares the summation result with the check code carried at the end of the original command using modulo-2 subtraction. If the difference is zero, the verification passes; otherwise, the parsing fails. After successful verification, the algorithm converts the binary encoding of the operation type field into a standardized operation type value by multiplying it by a weighting coefficient and summing the results. The operation types include shutdown, startup, and power adjustment. The algorithm multiplies the original byte sequence of the operation value field by the corresponding bit weight in big-endian order and sums the results to obtain an integer or floating-point value. For power adjustment, the result is the target power value; for startup and shutdown, it is the status flag. The algorithm multiplies the source priority level field by a preset level mapping factor to obtain the priority code corresponding to the local, field, or central control center. At the same time, the request timestamp field is obtained by adding the system base time to the relative time offset in the instruction to obtain the absolute timestamp. Finally, these parsed data are combined to form a preliminary parsed instruction, which is the output of this sub-step and is used for subsequent operator authentication and authorization verification. The original control instruction includes the target device identifier, operation type, operation parameters, source priority level, and request timestamp.

[0022] S12. Based on the preliminary parsing instructions, perform two-factor authentication and token verification to generate token authentication instructions. In this step, the string representations of the three fields are first concatenated into a continuous string in a fixed order. Then, the dynamic token key bound to the current operator's identity is read from the operator's pre-shared key storage area. The algorithm uses the hash message authentication code function to process the concatenated string: the dynamic token key is used as the key input, the concatenated string is used as the message input, and a fixed-length hash value is obtained through multiple iterations of compression and XOR operation. This hash value is compared with the expected hash value pre-stored in the system, and the bitwise XOR sum of the two is calculated. If the XOR result is zero, the authentication flag is true; otherwise, it is false. After the comparison is completed, the algorithm merges the authentication flag as additional information with the preliminary parsing instruction to form an authenticated instruction. If the authentication flag is true, the authenticated instruction retains all fields of the preliminary parsing instruction and is marked as passable; otherwise, the instruction is rejected and an invalid flag is added to the authenticated instruction, terminating the subsequent processing flow. The entire verification process combines two factors: the pre-shared key known to the operator and the real-time code generated by the dynamic token held by the operator. This ensures that only control instructions issued by legitimate operators can enter the subsequent permission screening stage, thereby improving the security of remote control of new energy equipment and preventing instruction forgery or replay attacks. The final output authenticated instruction contains the original parsing information and the authentication result.

[0023] S13. Perform a minimum permission screening on the token authentication command to obtain the parsed control command. In this step, the target device's unique identifier and operation type are used as a combined index to query the preset permission policy table to obtain the minimum permission level value required to execute the operation. Then, the algorithm subtracts the required minimum permission level value from the operator's role level value and calculates the difference. If the difference is greater than or equal to zero, the permission flag is set to one, indicating that the operator's permission meets the requirements. If the difference is negative, the permission flag is set to zero, indicating that the permission is insufficient. After the difference comparison is completed, the algorithm merges the permission flag as an additional field with all the information in the authenticated command to form the parsed control command. When the permission flag is zero, the system immediately terminates the subsequent control process and no longer executes the command issuance. When the permission flag is one, the parsed control command is completely preserved and passed to the next step.

[0024] This invention employs a state machine-based instruction decoding algorithm, which can resist buffer overflow and protocol tampering attacks, significantly enhancing the defense against malicious instructions and improving the robustness of communication protocols for new energy equipment. It utilizes a two-factor dynamic token verification algorithm to effectively prevent replay attacks, man-in-the-middle attacks, and offline password guessing attacks, significantly improving the authenticity and integrity of remote control instructions and strengthening the identity authentication strength of new energy power stations. Furthermore, it employs a role-based fine-grained permission matching algorithm. By querying a preset policy table, it obtains the minimum permission level required for a specific device's unique identifier and operation type combination, and then compares the difference between the operator's role level and this minimum permission level. This achieves independent permission control for each new energy device and each operation type, with finer granularity, preventing invalid instructions from entering subsequent core arbitration steps and reducing system overhead.

[0025] S2. Based on the parsed control instructions, multi-source queue candidate processing is performed to obtain candidate arbitration instructions. Conflict detection arbitration processing is then performed based on the candidate arbitration instructions to obtain priority arbitration instructions. Existing technologies cannot dynamically adjust priorities according to the real-time health status of the equipment and lack conflict type detection and resolution mechanisms. When multiple instructions are mutually exclusive or the power adjustment directions are opposite, fixed priorities cannot effectively adjudicate, which may lead to equipment malfunctions. The instruction waiting time factor is not considered. Instructions that have not been executed for a long time may be outdated, but fixed priorities still retain them, affecting control timeliness. To solve the above problems, the specific implementation steps are as follows: S21. Based on the parsed control instructions, a multi-source instruction set is obtained by extracting the status snapshot of the multi-source queue. In this step, all parsed control instructions from three sources—local, field, and central control center—are simultaneously retrieved from the local cache. For each source, the algorithm extracts its source priority level, operation type, operation value, and request timestamp value, and arranges these three field values ​​in a fixed order to form a multi-source instruction set containing three elements. At the same time, the algorithm reads three raw values ​​from the latest heartbeat packet from the device: device health score, available resource margin, and device-side emergency flag. The device health score is a decimal between zero and one, which remains unchanged after multiplying by one. The available resource margin is also multiplied by one to obtain its raw ratio. The device-side emergency flag is a value of zero or one, which is multiplied by one to obtain the flag itself. The algorithm then arranges these three product results in sequence to form the device's real-time status. Finally, the multi-source instruction set and the device's real-time status are used as the output of this sub-step, which is the multi-source instruction set, and passed to the subsequent dynamic priority weight calculation stage.

[0026] S22. Based on the multi-source instruction set, perform dynamic priority processing to obtain dynamic priority weights. In this step, firstly, calculate the source priority factor by dividing the source priority level value by two, then subtract the quotient from one, and multiply the difference by the first adjustable coefficient. Secondly, calculate the health factor by subtracting the device health score from one, and multiply the difference by the second adjustable coefficient. Thirdly, calculate the emergency flag factor by extracting the operation type of the instruction. If it is an emergency shutdown, the emergency flag value is one; otherwise, it is zero. Then, multiply the flag value by the third adjustable coefficient. Fourthly, calculate the waiting time factor by extracting the request timestamp of the instruction, subtracting the request time from the current time to obtain the waiting time in seconds, adding one to the waiting time, calculating its reciprocal, and multiplying the reciprocal by the fourth adjustable coefficient. After completing the calculation of these four products, the algorithm adds the four products together to obtain the dynamic priority weight value of the instruction. The adjustable coefficients employ a multi-objective optimization method based on historical data. First, real-time status data of multi-source command conflicts that actually occurred at new energy power plants over a period of time, along with corresponding equipment status data, are collected to form a training sample set containing input features and the optimal arbitration result. For each sample, four adjustable coefficients are considered as variables to be optimized, and an objective function is constructed. This function calculates the deviation between the arbitration result indicated by the dynamic priority weight obtained from the current coefficient combination and the correct result labeled by experts in the sample. The deviation is calculated by summing the squared differences in the ranking scores of all commands in each conflict case. Subsequently, the gradient descent algorithm is used to iteratively update the four coefficients. In each iteration, the partial derivative of the objective function with respect to each coefficient is calculated, and the new coefficient value is obtained by subtracting the learning rate multiplied by the partial derivative from the current coefficient value. At the same time, the constraint that the sum of the four coefficients is one is maintained. That is, after each update, the four new values ​​are normalized by dividing each coefficient by the sum of the four coefficients. The iteration continues until the change in the objective function is less than a preset threshold. The coefficient combination obtained at this point is the optimal adjustable coefficient.

[0027] S23. Arbitrate the dynamic priority weights and multi-source instruction set to output candidate arbitration instructions. In this step, the first instruction in the multi-source instruction set is set as the current candidate instruction, and its corresponding dynamic priority weight value is set as the current maximum weight. Then, the next instruction in the set and its weight value are taken out in turn. The current maximum weight is subtracted from the weight value of the instruction. If the difference is negative, it means that the weight of the instruction is greater. The current maximum weight is updated to the weight value of the instruction, and the current candidate instruction is replaced with the instruction. If the difference is positive, the current candidate instruction is kept unchanged. If the difference is zero, it means that the weights of the two instructions are equal. At this time, the request timestamps of the two need to be compared. In the timestamp comparison, the request timestamp value of the current candidate instruction is extracted and the request timestamp value of the instruction to be compared is compared. The difference between the candidate timestamp and the timestamp to be compared is calculated. If the difference is positive, it means that the timestamp of the instruction to be compared is smaller, i.e., earlier. The current candidate instruction is replaced with the instruction. If the difference is negative or zero, the original candidate is kept unchanged. The process continues until all instructions are processed. The last remaining instruction is the candidate arbitration instruction.

[0028] S24. Based on the candidate arbitration instructions, the conflict level is checked to obtain the conflict detection result. In this step, the total number of instructions in the multi-source instruction set is counted first, and this number is used as the denominator. Then, each other instruction in the set except for the candidate arbitration instructions is traversed, and the operation type value of the candidate arbitration instruction and the operation type value of the instruction are extracted. For start and stop operations, the two are compared with the stop flag and the start flag respectively. If the candidate is start and the other is stop, or the candidate is stop and the other is start, then the instruction is marked as a conflict instruction. For power adjustment operations, the power adjustment directions of the two instructions are compared. The operand value of the candidate instruction is calculated and then subtracted from the operand value of the other instruction. If the difference is positive and the candidate increases power while the other decreases power, or if the difference is negative and the candidate decreases power while the other increases power, then the instruction is marked as a conflicting instruction. The number of all marked conflicting instructions is counted, and this number is divided by the total number of instructions to obtain the quotient, which is used as the conflict level value. At the same time, the conflict type is determined according to the opposing types present in the conflicting instructions. If there is a mutually exclusive pair of start and stop, the conflict type is mutual exclusion conflict. If there is only a pair of opposite power adjustment directions, the conflict type is directional conflict. If there are no conflicting instructions, the conflict type is no conflict. Finally, the conflict type and conflict level are combined to form the conflict detection result and output.

[0029] S25. Based on the conflict detection results and candidate arbitration instructions, perform conflict resolution instruction processing to obtain the resolved instructions. In this step, first determine whether the conflict type in the conflict detection results is conflict-free. If so, directly output the candidate arbitration instructions as the resolved instructions. If there are mutual exclusion conflicts or directional conflicts, it is necessary to calculate the comprehensive score for each instruction in the multi-source instruction set. For each instruction, first extract the dynamic priority weight value of the instruction, and at the same time obtain the conflict level value from the conflict detection results. Calculate the difference between one and the conflict level, and multiply the difference by the dynamic priority weight to obtain the product A. Then, it is determined whether the instruction will cause the device to enter a protection state. If so, the security mode factor is set to 0.2; otherwise, it is set to 1. The security mode factor is multiplied by a preset security coefficient, which is a constant between zero and one, to obtain product B. Finally, product A is added to product B to obtain the comprehensive score of the instruction. The above calculation process is repeated for each instruction in the multi-source instruction set. After obtaining the comprehensive scores of all instructions, the instruction corresponding to the maximum score is found and output as the resolved instruction. If multiple instructions have the same comprehensive score and are all the maximum values, the instruction with the earliest request timestamp is selected as the resolved instruction.

[0030] S26. Perform priority arbitration on the resolved instruction to obtain the priority-arbitrated instruction. In this step, firstly, extract the weight value corresponding to the resolved instruction from the previously calculated dynamic priority weight and use it as the final priority value. Then, set the conflict resolution flag to a value of one, indicating that the conflict has been successfully resolved. Subsequently, obtain the current system time as the arbitration timestamp. This timestamp is usually expressed as the number of milliseconds or seconds accumulated from a certain base time point. The final priority value, the conflict resolution flag value, and the arbitration timestamp value are regarded as three additional data fields. During the encapsulation process, the original fields in the deconstructed instruction, including target device identifier, operation type, operation value, source priority level, request timestamp, authentication flag, and permission flag, are concatenated with these three additional fields in a fixed order. Specifically, the total length of the original fields is calculated, and then the final priority value is multiplied by one and appended to the end, the conflict resolution flag is multiplied by one and appended, and the arbitration timestamp value is multiplied by one and appended to the end. The addition of these three values ​​does not change any of the original field values; it only forms a new data structure through append operations. Finally, the algorithm outputs a priority arbitration instruction, which contains complete original control information, authentication and permission verification results, arbitration decision basis, and conflict resolution metadata.

[0031] This invention not only considers the source priority level, but also introduces three dynamic factors: equipment health score, emergency flag, and instruction waiting time. The comprehensive weight is obtained by weighted summation through adjustable coefficients. When the temperature of a wind turbine bearing rises abnormally, the local control panel does not operate in time, while the central control center issues an emergency shutdown instruction. When a fixed priority is used, the local instruction still takes priority, causing the emergency shutdown to be blocked. When the equipment health score is extremely low and multiple sources issue emergency shutdowns, the weight calculation may result in multiple instructions having similar weights. This solution gradually increases the weight of instructions that have been waiting longer, avoiding indefinite backlog of instructions. It not only realizes dynamic priority arbitration, but also links with the subsequent equipment health prediction model, thus improving the overall response speed of new energy power plants to sudden failures.

[0032] This invention first identifies mutually exclusive and directional conflicts using a command collision detection algorithm, then calculates the comprehensive score of each command. A safety mode factor is introduced, significantly reducing the score if a command causes the device to enter a protection state. When an energy storage system simultaneously receives two conflicting commands—one charging at 500kW and the other discharging at 300kW—a fixed priority might incorrectly select the higher-level charging command while ignoring the potential over-discharge risk of the discharging command, or conversely, cause overcharging. When the conflict level is high and all commands could potentially lead to device protection, existing technologies lack a joint trade-off mechanism between safety and priority. They cannot select a relatively safe command from multiple dangerous commands, nor do they introduce a safety mode factor to quantify the impact of commands on the device's protection state. Furthermore, they do not set a safety coefficient to limit the upper limit of the safety factor's contribution to avoid over-conservatism. They also do not record the conflict level, making it impossible to perceive the severity of the conflict. This method not only solves the conflict resolution problem but also links with subsequent corrections. The resolved command is accompanied by a conflict resolution flag, allowing corrections to adjust the deviation tolerance based on this flag. This reduces the number of corrections without sacrificing safety, thus improving control efficiency.

[0033] This invention introduces a metadata encapsulation algorithm to achieve full traceability of the arbitration process. After priority arbitration, the final priority value, conflict resolution flag, and arbitration timestamp are appended to the instruction. This metadata is transmitted to subsequent steps along with the original instruction. For example, if a photovoltaic power station experiences an accidental shutdown, maintenance personnel cannot reconstruct how the dynamic priority weights of each instruction were calculated from historical records, nor can they determine which instruction was selected in the conflict resolution process and the basis for it. When multiple stations simultaneously issue instructions to the same device, existing technologies do not link arbitration metadata with audit logs, resulting in the inability to form an immutable chain of evidence in the instruction game process between multiple control sources, making subsequent liability determination difficult. This solution, through the conflict resolution flag in the metadata, can link to subsequent security audits, automatically triggering blockchain evidence storage to form an immutable chain of arbitration evidence. This not only achieves transparency in arbitration but also links with auditing, using arbitration metadata directly as input for digital signatures, enhancing the compliance and non-repudiation of the entire process.

[0034] S3. Perform health satisfaction analysis on the priority arbitration instructions to obtain the equipment health feature vector and resource satisfaction. Based on the equipment health feature vector and resource satisfaction, perform verification deviation processing to obtain the pre-verification pass instructions. Existing technology only compares the current health score of the equipment with a fixed threshold, which cannot predict the amount of equipment health degradation after the instruction is executed. This leads to old equipment being permanently rejected from light-load operation because the health score is continuously lower than the threshold, resulting in resource waste. It only considers a single factor without taking into account multiple constraints such as maximum allowable power, ambient temperature, and irradiance. It is prone to misjudgment in extreme environments or power limits. The verification threshold is fixed and cannot adapt to the increasing trend of prediction error caused by equipment performance degradation. When the prediction model deviation increases after the equipment ages, it is easy to cause a large number of valid instructions to be rejected or dangerous instructions to be missed. To solve the above problems, the specific steps are as follows: S31. After priority arbitration, the instruction is processed through health feature parameters to obtain the device health feature vector and resource parameter vector. In this step, the unique identifier of the target device is first extracted from the instruction, and a real-time data reading request is sent to the device IoT gateway based on the identifier. For the device health feature vector, the core temperature value, vibration amplitude value, cumulative start-stop count value and capacity decay rate value are obtained in sequence. Each value is standardized. The core temperature value is multiplied by one and then zero is added to obtain the temperature feature value. The vibration amplitude value is multiplied by one and then zero is added to obtain the vibration feature value. The cumulative start-stop count value is first divided by 10,000 for normalization, and then the quotient is multiplied by one and then zero is added to obtain the count feature value. The capacity decay rate value is multiplied by one and then zero is added to obtain the decay feature value. These four calculation results are arranged in order to form the device health feature vector. For the resource parameter vector, the energy storage state of charge (SOC), maximum allowable power, ambient temperature, and photovoltaic irradiance are obtained sequentially. The SOC is multiplied by one and remains unchanged. The maximum allowable power is first normalized by dividing by the rated power of the equipment, and then multiplied by one to obtain the normalized power limit. The ambient temperature is multiplied by one and then zero is added to obtain the ambient temperature value. The photovoltaic irradiance is first scaled by one thousand, and then multiplied by one to obtain the normalized irradiance value. These four calculation results are arranged in order to form the resource parameter vector. Finally, the equipment health feature vector and the resource parameter vector are output together.

[0035] S32. Based on the equipment health feature vector and priority arbitration instructions, process the equipment health degradation amount to obtain the predicted health degradation amount. In this step, firstly, extract four values ​​from the equipment health feature vector: core temperature, vibration amplitude, cumulative start-stop count, and capacity decay rate. Simultaneously, extract the operation type value and operation value from the priority arbitration instructions. During the calculation of the Long Short-Term Memory network, the algorithm first processes the forget gate, multiplying the four health feature values ​​by the first set of weight coefficients of the forget gate and summing them. Then, multiply the operation type value by the operation type weight of the forget gate, and multiply the operation value by the operation value weight of the forget gate. After adding the sum, the bias term of the forget gate is added to obtain the value before the forget gate is activated. This value is compressed to between zero and one by the Sigmoid function and used as the output of the forget gate. Then the input gate is processed. The four health feature values ​​are multiplied by the first set of weight coefficients of the input gate and summed. Then the weight term of operation type and operation value is added. After adding the input gate bias, the output of the input gate is obtained by the Sigmoid function. The weight coefficients actually correspond to the independent weight matrices of the forget gate, input gate, candidate memory unit and output gate, respectively. They can be obtained by the training algorithm based on time backpropagation, which will not be elaborated here. Simultaneously, candidate memory units are calculated. The health feature values ​​are multiplied by the first set of weight coefficients of the candidate units and summed. The weighted terms of operation type and operation value are added, and the candidate unit bias is added. The result is then compressed to between -1 and 1 using the hyperbolic tangent function to obtain the candidate memory value. The memory unit state of the previous time step is multiplied by the forget gate output to obtain the retained historical information. The input gate output is then multiplied by the candidate memory value to obtain the new information. The retained historical information and the new information are added to update the memory unit state of the current time step. Finally, the output gate is processed. The health feature values ​​are multiplied by the first set of weight coefficients of the output gate and summed. The weighted terms of operation type and operation value are added, and the result is then passed through the Sigmoid function to obtain the output gate output. The current memory unit state is transformed by the hyperbolic tangent function and multiplied by the output gate output to obtain the current hidden state. This hidden state passes through a fully connected layer. Each component of the hidden state is multiplied by its corresponding output weight and summed. The output bias is added, and the result is then input into the Sigmoid function. Finally, a value between zero and one is obtained, which is the predicted health degradation amount.

[0036] S33. Perform resource condition satisfaction analysis on the resource parameter vector and priority arbitration instructions to obtain resource satisfaction. In this step, first extract the operation type value from the priority arbitration instructions to determine whether it is a power regulation instruction or a start / stop instruction. If it is a power regulation instruction, extract the energy storage state of charge value and the maximum allowable power value from the resource parameter vector. At the same time, extract the absolute value of the operation value from the instruction as the target power absolute value. Divide the energy storage state of charge value by the preset minimum state of charge threshold required to perform regulation to obtain the first ratio. Divide the maximum allowable power value by the absolute value of the target power to obtain a second ratio. Calculate these two ratios separately, take the smaller value, and compare it with the first value. Take the minimum of the three as the resource satisfaction level. If the first ratio is less than the second ratio and less than one, the resource satisfaction level is equal to the first ratio. If the second ratio is less than the first ratio and less than one, the resource satisfaction level is equal to the second ratio. If both ratios are greater than or equal to one, the resource satisfaction level is equal to one. If it is a start / stop command, extract the energy storage state of charge value and the maximum allowable power value from the resource parameter vector. Compare the energy storage state of charge value with the preset shutdown protection threshold. If the energy storage state of charge value is greater than the shutdown protection threshold and the maximum allowable power value is greater than zero, the resource satisfaction level is assigned one; otherwise, the resource satisfaction level is assigned zero. Regardless of the command type, the final output resource satisfaction level is a value between 0 and 1.

[0037] S34. Based on the device health feature vector and resource parameter vector, a historical deviation sequence is generated by checking the deviation sequence. In this step, the unique identifier of the device is first extracted from the device health feature vector, and the historical prediction deviation record related to the device is searched in the local cache according to the identifier. For each historical moment, the algorithm reads the actual health value and the predicted health value at that moment, calculates the absolute value of the difference between the two, subtracts the predicted health value from the actual health value to get the difference, and if the difference is negative, takes its opposite number to get the prediction deviation at that moment. A fixed window length value is preset, for example, the window length value is ten. Then, the prediction deviation values ​​of the ten most recent moments before the current moment are retrieved from the cache and arranged in chronological order from earliest to latest to form a sequence containing ten elements. If the historical records in the cache are less than ten moments, the missing positions are filled with zero values ​​to ensure that the sequence length is always equal to the window length. During the arrangement process, each deviation value is not subjected to any multiplication, division or addition and subtraction transformations and is directly used as an element of the sequence. Finally, the algorithm outputs the historical deviation sequence, where each element represents the absolute error between the actual health status and the predicted health status at the corresponding moment.

[0038] S35. Based on the historical deviation sequence, adaptive threshold processing is performed to generate an adaptive threshold factor. In this step, the latest predicted deviation value at the current time and the exponentially weighted moving average calculated at the previous time are first extracted from the sequence. If the moving average at the previous time does not exist, it is initialized to zero. The algorithm multiplies the smoothing factor by the latest predicted deviation value to obtain a product C. The smoothing factor is subtracted from the moving average at the previous time to obtain a product D. The product C and the product D are added to obtain the exponentially weighted moving average at the current time. This calculation makes the influence of recent deviations on the average value greater, and the influence of long-term deviations decays exponentially. After the moving average calculation is completed, the scaling factor is multiplied by the moving average to obtain a product E. The baseline threshold is then added to the product E to obtain the initial adaptive threshold factor. The algorithm then limits the range of this initial value: if the value is less than 0.05, it is assigned the value of 0.05; if the value is greater than 0.3, it is assigned the value of 0.3; if it is between the two, it remains unchanged. The final output adaptive threshold factor is a value between 0.05 and 0.3. This value will be used in the subsequent pre-verification decision to compare with the predicted health degradation. When the historical deviation sequence shows an increasing trend, the exponentially weighted moving average increases accordingly, causing the adaptive threshold factor to rise, thereby widening the allowable deviation range for verification. Conversely, the threshold is tightened to achieve dynamic adjustment of the verification standard.

[0039] S36. Perform comprehensive verification processing on the adaptive threshold factor, predicted health degradation amount, and resource satisfaction to obtain the pre-verification pass instruction; In this step, firstly, compare the value of the predicted health degradation amount with the value of the adaptive threshold factor, and subtract the predicted health degradation amount from the adaptive threshold factor. If the difference is positive, it means that the predicted degradation amount is less than the threshold and the health condition is met. If the difference is negative or zero, it is not met. At the same time, compare the value of the resource satisfaction with 0.8, and subtract 0.8 from the resource satisfaction. If the difference is positive, it means that the resource satisfaction is greater than 0.8 and the resource condition is met. Otherwise, it is not met. When the health condition and the resource condition are met at the same time, the verification flag is assigned a value of one. If either condition is not met, the verification flag is assigned a value of zero. After assigning the flag value, all original fields are extracted from the priority arbitration instruction. The verification flag, predicted health degradation value, resource satisfaction value, adaptive threshold factor value, and current timestamp are added as five additional fields and appended to the end of the original fields in sequence. The specific appending process is as follows: the verification flag is multiplied by one and then concatenated, the predicted health degradation value is multiplied by one and then concatenated, the resource satisfaction value is multiplied by one and then concatenated, the adaptive threshold factor is multiplied by one and then concatenated, and the current timestamp is multiplied by one and then concatenated to form the pre-verification pass instruction. If the verification flag is zero, an alarm message is generated at the same time as the pre-verification pass instruction is output, and the subsequent control process is terminated. If the verification flag is one, the pre-verification pass instruction is completely output to the next core step. The final output pre-verification pass instruction contains complete original instruction information, arbitration information, and various judgment bases of the pre-verification.

[0040] This invention employs a Long Short-Term Memory (LSTM) network, using core device temperature, vibration amplitude, cumulative start-stop count, and capacity decay rate as input features. Combined with operation type and operation value, it predicts the dimensionless degradation increment after executing the instruction. This not only achieves a quantitative assessment of the long-term impact of the instruction but also adaptively links with historical prediction errors, dynamically adjusting the tolerance when the model accuracy fluctuates. Compared to a fixed threshold, this improves the availability of older equipment and significantly reduces the rate of missed dangerous instructions. For example, when equipment is in the middle of performance degradation, the prediction model itself may have a large bias. Equipment with a low health score but actually performing light-load operations, resulting in only a small degradation increment, is mistakenly rejected, causing a waste of available resources. Conversely, equipment with a high health score but a sudden increase in degradation after execution due to local anomalies is mistakenly released, posing a safety hazard. This invention not only integrates energy storage SOC and power limits but also links with environmental monitoring. It automatically rejects over-limit commands in extreme weather or equipment derating scenarios, significantly reducing resource-related operational errors. It not only achieves adaptive threshold adjustment but also mutually reinforces the LSTM prediction model. The historical deviation sequence records the LSTM prediction error, and the adaptive threshold dynamically adjusts based on this error, ensuring the verification standard matches the model's current accuracy. This significantly reduces the false rejection rate while maintaining a high interception rate for dangerous commands. For example, in the early stages of a photovoltaic inverter's operation, the prediction deviation sequence is stable. After five years, due to aging, the prediction deviation increases. Although parameters such as ambient temperature and irradiance are not directly used as verification conditions, they indirectly affect resource satisfaction by influencing power limits. Existing technologies cannot automatically trigger verification failures and alarms during extreme weather or equipment derating. When a sudden equipment failure causes a sharp drop in actual health values, the prediction deviation increases dramatically. Existing fixed thresholds cannot quickly respond to such sudden changes. If the threshold is too strict, it will falsely reject emergency commands; if the threshold is too lenient, it may miss dangerous commands. Furthermore, the lack of a dynamic balancing mechanism prevents correct rejection during sudden failures.

[0041] This invention compares the predicted degradation amount with the adaptive threshold and the resource satisfaction level through comprehensive judgment. Only when both are satisfied does the verification pass. This not only completes the pre-verification but also provides a complete decision basis for subsequent auditing and re-arbitration. It realizes cross-step data linkage and enhances the interpretability and fault tolerance of the entire control system. For example, in a multi-source control environment, an instruction from one source is rejected due to resource verification failure, but instructions from other sources may have different operating parameters, such as a lower target power value, and can meet the current resource constraints. Due to the lack of a linkage mechanism to trigger re-arbitration and select alternative instructions after verification failure, the overall control success rate decreases, and the redundancy advantage of multi-source control cannot be fully utilized.

[0042] S4. Based on pre-verification, high-precision delay processing is performed through instructions to obtain the instantaneous deviation vector and the measured response delay. The steady-state deviation correction accuracy processing is then performed on the instantaneous deviation vector and the measured response delay to obtain the execution feedback result. Existing technologies lack millisecond-level continuous feedback sampling, making it impossible to obtain the dynamic deviation trajectory during the device response process. This results in a lack of awareness of transient problems such as overshoot and oscillation. When the actual output has a steady-state error compared to the target value due to device aging, environmental changes, or communication delay jitter, it cannot be automatically compensated. It does not consider the dynamic changes in response delay and cannot distinguish whether it is due to device failure or improper control parameters. To solve the above problems, the specific steps are as follows: S41. Based on the pre-verification pass instruction, perform high-precision instruction processing to obtain the instruction issuance record and initial device state. In this step, first obtain the value of the current system time and use this value as the instruction issuance time. Then, read the theoretical response delay value corresponding to the device from the device parameter table, add the issuance time value to the theoretical response delay value to obtain the expected response time value. At the same time, read the current power value and start / stop status value of the device through the telemetry interface. Multiply the power value by one and add zero to obtain the initial power value. Multiply the start / stop status value by one and add zero to obtain the initial status value. Combine all fields in the pre-verification pass instruction with the issuance time value and the expected response time value to form the instruction issuance record. Combine the initial power value and the initial status value to form the initial device state. Finally, output the instruction issuance record and the initial device state together.

[0043] S42. Based on the instruction issuance record, a real-time feedback sequence is generated through continuous acquisition and feedback processing. In this step, the value of the instruction issuance time is first extracted from the instruction issuance record. The sampling period value is multiplied by 100 milliseconds to obtain the sampling interval. The maximum number of sampling points is multiplied by 100 to obtain the total number of samples. An empty sequence is initialized to store the sampling results. The current sampling sequence number is set to start from 1. For each sampling sequence number, the sampling time is first calculated. The sampling sequence number is subtracted by 1 to obtain the offset sequence number. The offset sequence number is multiplied by the sampling interval to obtain the offset time. The instruction issuance time value is added to the offset time to obtain the current sampling time value. Subsequently, the actual power value and actual start / stop status value of the device at that sampling moment are read through the telemetry interface. The actual power value is multiplied by one and then zero is added to obtain the power value at that moment. The actual start / stop status value is multiplied by one and then zero is added to obtain the status value at that moment. The sampling moment value, power value, and status value are combined into a sampling point and added to the end of the sequence. The sampling number is incremented by one. The above process is repeated until the sampling number exceeds the maximum number of sampling points. Finally, a real-time feedback sequence containing one hundred sampling points is output. Each sampling point consists of three data points: sampling moment, actual power, and actual status, which are used for subsequent deviation calculation and response delay measurement.

[0044] S43. Perform control deviation delay processing on the real-time feedback sequence and the pre-verification instruction to obtain the instantaneous deviation vector and the measured response delay. In this step, each sampling point is first extracted sequentially from the real-time feedback sequence. Each sampling point includes the sampling time, actual power value, and actual state value. At the same time, the operation type value and target operation value are extracted from the pre-verification instruction. For power adjustment instructions, the target power is extracted, and for start / stop instructions, the target state is extracted. If the operation type is power adjustment, the actual power value is subtracted from the target power value. If the difference is negative, the opposite number is taken to obtain the absolute value of the instantaneous power deviation of the sampling point. If the operation type is start / stop, the actual state value is subtracted from the target state value. If the difference is negative, the opposite number is taken to obtain the absolute value of the instantaneous state deviation. Arrange the absolute values ​​of the deviations of all sampling points in the sampling order to form an instantaneous deviation vector. Then calculate the measured response delay. For each sampling point in the real-time feedback sequence, first calculate the absolute value of the instantaneous power deviation at that point. Divide the absolute value of the deviation by the target power value to obtain the relative deviation value. Sequentially determine whether the relative deviation value of each sampling point is less than or equal to 0.02. When the first sampling point that meets the condition is found, extract the sampling time value of that sampling point and use it as the measured response delay. If there is no sampling point that meets the condition in the entire sequence, set the measured response delay to an invalid value. Finally, output the instantaneous deviation vector and the measured response delay.

[0045] S44. Based on the instantaneous deviation vector, the steady-state predicted deviation is obtained through steady-state deviation trend processing. In this step, the deviation value of the last sampling point is first extracted from the instantaneous deviation vector as the deviation value of the previous moment. Then, the pre-calibrated state transition coefficient is read and multiplied by the deviation value of the previous moment to obtain the first product. At the same time, the control input coefficient is read and multiplied by the current control quantity value, such as the pulse width modulation duty cycle, to obtain the second product. The first product and the second product are added to obtain the preliminary prediction value. Finally, the process noise value is added to the preliminary prediction value to obtain the final steady-state predicted deviation. The process noise value is a small constant pre-set according to the equipment operating state. If the control quantity value is unavailable at the current moment, the second product is set to zero. This algorithm predicts the expected deviation value of the equipment in steady state through the linear combination of state transition and external control input.

[0046] S45. Compensation and correction are performed based on the steady-state prediction deviation and the measured response delay to obtain the compensation and correction amount. In this step, the proportional term is first calculated by multiplying the proportional coefficient by the steady-state prediction deviation value to obtain the first product. Then, the integral term is calculated by multiplying the integral coefficient by the cumulative integral value E of the steady-state prediction deviation, expressed as follows: In this product, SSPD represents the steady-state prediction deviation, dt represents the time infinitesimal element, and the second product is obtained. Then, the differential term is calculated, and the differential coefficient is multiplied by the rate of change of the steady-state prediction deviation. The rate of change is equal to the difference between the current deviation and the deviation at the previous moment divided by the time interval, and the third product is obtained. Then, the time delay compensation term is calculated. The theoretical response time delay value of the device is read, and the measured response time delay value is subtracted to obtain the time delay difference value. The time delay compensation coefficient is multiplied by this difference value to obtain the fourth product. The first, second, third, and fourth products are added together to obtain the compensation correction amount. If the correction amount is positive and the operation type is power regulation, it indicates that an additional positive power regulation amount needs to be added. If it is negative, it indicates that an additional reverse power regulation amount needs to be added. If it is a start-stop operation and the correction amount exceeds the preset threshold, a retry flag is output. The final output compensation correction amount is used to generate subsequent secondary correction commands. The proportional coefficient, integral coefficient, and derivative coefficient are pre-calculated using engineering tuning methods such as Ziegler-Nichols. The time delay compensation coefficient is obtained by fitting the linear relationship between the difference and the compensation coefficient using the least squares method. The proportional coefficient, integral coefficient, derivative coefficient, and time delay compensation coefficient are then finely adjusted step by step using the gradient descent algorithm. This process will not be elaborated here.

[0047] S46. Perform secondary correction and accuracy processing on the compensation correction amount and the pre-verification pass instruction to obtain the execution feedback result. In this step, the target power value is first extracted from the pre-verification pass instruction, and the target power value is multiplied by 0.01 as the deviation threshold that does not need correction. The absolute value of the steady-state prediction deviation is compared with the threshold. If the absolute value of the steady-state prediction deviation is less than the threshold, no correction is needed and the process jumps directly to the output stage. Otherwise, the correction process is entered. In the correction process, the operation value in the pre-verification pass instruction is added to the compensation correction amount to obtain a new operation value, and a correction instruction is generated and sent to the device. Then, wait for the next round of feedback sampling, recalculate the new steady-state prediction deviation, and repeat the above comparison and correction process. After each correction, recalculate the product of the current absolute value of the deviation and the target power multiplied by 0.001 and compare it. Stop iterating when the current absolute value of the deviation is less than the product. Record the number of iterations as the number of corrections. Finally, the algorithm extracts the original instruction information from the pre-verification pass instruction, obtains the final power value and the final state value from the last feedback, and combines the number of corrections, the accuracy set to 100%, and the current timestamp to form the execution feedback result and output it.

[0048] This invention achieves millisecond-level continuous sampling and dynamic deviation calculation, solving the problem of the invisibility of the response process in open-loop control. It can completely capture dynamic processes such as power ramp-up, overshoot, and oscillation. Overshoot is detected through the sampling sequence, triggering closed-loop correction. The instantaneous deviation vector can identify the dead zone where the power remains near the target value for a long time. At this time, the steady-state deviation predicted by the Kalman filter is non-zero, thus triggering compensation correction. It not only realizes the visualization of the response process but also links with PID correction. The measured response delay is directly used for the delay compensation term, allowing the correction amount to predict the lag in advance, significantly reducing the overshoot. At the same time, it greatly... To significantly shorten response time, for example, when a photovoltaic inverter executes a command to adjust the power from 100kW to 500kW, the open-loop control waits 2 seconds and reads the power as 498kW, considering it a success. However, in actual operation, the power overshoot reached 530kW, triggering protection. When the device has a response dead zone, such as when a small signal is not responded to, the open-loop control only reads the final steady-state value and lacks millisecond-level continuous sampling, making it unable to detect overshoot or oscillation during power adjustment. Furthermore, when the device has a response dead zone, traditional methods cannot detect the minute deviations caused by the lack of response to small signals, which may trigger device protection or leave safety hazards.

[0049] This invention constructs a dual-model correction system combining Kalman filtering and PID lead-lag to address the problem of unautomatic elimination of steady-state errors. It uses PID with time delay compensation to generate the correction value and iterates until the final deviation is significantly smaller than the target value. A time delay compensation term is introduced, dynamically adjusting the correction value based on the difference between the measured response delay and the theoretical delay. This is linked to the measured response delay; a large delay difference increases the compensation coefficient, compensating for lag in advance, while a small delay difference decreases the correction value, preventing over-adjustment. This not only eliminates static errors but also adapts to changes in communication delay, making it suitable for hybrid networking scenarios. The number of calibration commands is significantly reduced, and the calibration failure rate caused by time delay jitter is greatly reduced, which significantly improves control efficiency and robustness. For example, a certain energy storage PCS has a 5kW steady-state error between the actual power and the command due to temperature changes. Open-loop control cannot correct this. When the communication network time delay jitters randomly, such as from 20ms to 200ms, open-loop control cannot eliminate the steady-state error caused by environmental changes such as temperature drift. Existing technologies lack the ability to adaptively compensate for the random jitter of communication network time delay, which leads to over-adjustment of the calibration amount and causes power oscillation.

[0050] This invention employs iterative convergence judgment and correction count recording to achieve a verifiable closed loop with extremely high accuracy. By progressively tightening the threshold, it ensures that the error decays exponentially after each correction. This not only guarantees extremely high accuracy under normal operating conditions but also enables rapid identification and isolation of faults when equipment malfunctions, reducing fault location time from hours to minutes. Furthermore, the statistical analysis of correction counts provides quantitative indicators for equipment health assessment, achieving integrated control and diagnosis. In practice, a wind farm batch-started and shut down 50 wind turbines, with 3 turbines not completely shutting down due to response delays. When equipment experiences severe deviations in response due to hardware failure, such as power stalling, existing iterative correction technologies fail to converge. Batch control cannot guarantee accurate response for each device, leading to omissions due to individual response delays. Moreover, when hardware failures cause severe deviations such as power stalling, open-loop control cannot identify the fault type, and iterative correction may loop infinitely or be misjudged as successful, lacking an abnormal convergence mechanism and fault alarm.

[0051] S5. Based on the execution feedback results, the audit log is obtained through multi-source audit processing. Existing technology lacks dual redundancy verification of power and status, and cannot detect misjudgments caused by abnormal telemetry data or transient interference. Log records are easily tampered with or deleted internally, and cannot provide irrefutable evidence when tracing back afterward. To solve the above problems, the specific implementation steps are as follows: S51. Based on the execution feedback results, a consistency comparison report is generated through multi-source result consistency comparison. In this step, the final power value, target power value, final state value, and target operation type value are first extracted from the execution feedback results. The target power value is multiplied by 0.001 to obtain the error tolerance value. The difference between the final power value and the target power value is calculated. If the difference is negative, its opposite is taken to obtain the absolute deviation value. The absolute deviation value is compared with the error tolerance value. If the absolute deviation value is less than the error tolerance value, the power matching flag is set to one; otherwise, it is set to zero. At the same time, the final state value is compared with the target operation type value. If the two are equal, the state matching flag is set to one; otherwise, it is set to zero. The power matching flag and the state matching flag are combined to form a consistency comparison report.

[0052] S52. Based on the consistency comparison report, perform digital non-repudiation signature processing to obtain a signed record. In this step, firstly, all fields in the consistency comparison report and all fields in the execution feedback result are concatenated in a fixed order to form a continuous long string. Then, the national cryptographic SM3 hash function is used to process the long string, dividing it into several 512-bit groups. Each group is iteratively compressed and XORed to obtain a hash value of a fixed length. Next, the operator's or system's private key is read from the key management module, and the hash value is signed using the national cryptographic SM2 signature algorithm. A random number is generated, and the coordinates of the point on the elliptic curve are calculated. The first and second components of the signature value are obtained through modulo and multiplication operations. The two components are combined to form the final digital signature. At the same time, the current system time is obtained as the signature time. The original consistency comparison report, execution feedback result, calculated digital signature, and signature time are combined to form a signed record. The national cryptographic SM2 signature algorithm is an asymmetric digital signature algorithm based on elliptic curve cryptography issued by the State Cryptography Administration of China, which will not be elaborated here.

[0053] S53. Perform blockchain auditing on the signed records to obtain the audit log. In this step, the Merkle root hash value of the previous data block is first read from local storage as the basis for this construction. The current signed record is concatenated with the previous Merkle root hash value to form a string of fixed length. Then, a hash operation is performed on the string to divide it into several fixed-length groups. Each group is iteratively compressed and XORed to obtain the Merkle root hash value of the current data block. At the same time, the current value of the global block counter is read and incremented by one to obtain the new block height value. The algorithm combines the signed record, the calculated Merkle root hash value, and the block height value together to form the audit log. Finally, the audit log is broadcast to all audit nodes for distributed storage. The entire process does not involve multiplication or division operations and is mainly completed through string concatenation, hash compression, and addition operations.

[0054] This invention addresses compliance and non-repudiation issues through non-repudiation digital signatures based on the national cryptographic algorithms SM2 and SM3. It concatenates the consistency comparison report and execution feedback results, calculates the hash value using SM3, and then signs it with an SM2 private key, forming a signed record containing the complete control chain. Any node can verify the signature's authenticity using its public key, eliminating the need for a centralized auditor and further enhancing trust in cross-domain auditing. The Merkle tree storage algorithm is used to write the blockchain audit log, resolving issues of log tampering and single points of failure. This achieves end-to-end traceability from instruction initiation to final audit. In special scenarios such as hidden fault identification, cross-domain trust, and batch verification, it demonstrates comprehensive advantages of high security, high compliance, and high efficiency, providing a complete trust system for remote control of new energy equipment.

[0055] Example 2: Because current technologies are unaware of transient issues such as overshoot and oscillation, steady-state errors cannot be automatically compensated, dynamic changes in response delay are not considered, and specific problems cannot be accurately identified, please refer to [the relevant documentation / reference]. Figure 2 The diagram shown is a structural block diagram of a remote control and security verification system for new energy equipment provided in this embodiment. The system includes a parsing module, a priority arbitration module, a pre-verification module, an execution feedback module, and an audit module. The parsing module is used to collect the original control commands of the target device and obtain the parsed control commands by parsing verification and initial permission screening based on the original control commands. The priority arbitration module is used to process multi-source queue candidates based on parsed control instructions to obtain candidate arbitration instructions, and to perform conflict detection arbitration processing based on the candidate arbitration instructions to obtain priority arbitration instructions. The pre-verification module is used to perform health satisfaction analysis on the instructions after priority arbitration, obtain the equipment health feature vector and resource satisfaction, and obtain the pre-verification pass instruction based on the equipment health feature vector and resource satisfaction through the verification deviation verification process. The execution feedback module is used to perform high-precision delay processing based on the pre-verification through instructions to obtain the instantaneous deviation vector and the measured response delay. The instantaneous deviation vector and the measured response delay are then processed for steady-state deviation correction accuracy to obtain the execution feedback result. The audit module is used to obtain audit logs through multi-source audit processing based on the execution feedback results.

[0056] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code, including but not limited to disk storage, CD-ROM, optical storage, etc.

[0057] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A new energy equipment remote control and security verification method, characterized in that, The steps of this method are as follows: collect the original control commands of the target device, and obtain the parsed control commands by parsing, verifying and initially screening permissions based on the original control commands; Based on the parsed control instructions, multi-source queue candidate processing is performed to obtain candidate arbitration instructions. Conflict detection arbitration processing is then performed based on the candidate arbitration instructions to obtain priority arbitration instructions. Health satisfaction analysis is performed on the priority arbitration instructions to obtain the equipment health feature vector and resource satisfaction. Based on the equipment health feature vector and resource satisfaction, the pre-verification pass instructions are obtained through the verification deviation process. Based on the pre-verification, high-precision time delay processing is performed through instructions to obtain the instantaneous deviation vector and the measured response time delay. Steady-state deviation correction accuracy processing is then performed on the instantaneous deviation vector and the measured response time delay to obtain the execution feedback result. Based on the execution feedback results, audit logs are obtained through multi-source auditing. Among them, high-precision delay processing based on pre-verification through instructions includes: high-precision processing of instructions based on pre-verification to obtain instruction issuance records and initial device status; Based on the instructions issued, the records are continuously collected and processed to generate a real-time feedback sequence; The real-time feedback sequence and pre-verification are processed by control deviation delay through instructions to obtain the instantaneous deviation vector and the measured response delay.

2. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, The process involves parsing, verifying, and initially screening permissions based on the original control instructions, including: parsing and format verification of the original control instructions to obtain preliminary parsed instructions; Based on the initial parsing instructions, perform two-factor authentication and token verification to generate token authentication instructions; The token authentication command is initially screened for least privilege to obtain the parsed control command.

3. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, Multi-source queue candidate processing based on parsed control instructions includes: extracting a multi-source instruction set from a multi-source queue state snapshot based on the parsed control instructions; Dynamic priority weights are obtained by performing dynamic priority processing based on multi-source instruction sets. Arbitration output candidate processing is performed on the dynamic priority weight and multi-source instruction set to obtain candidate arbitration instructions.

4. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, The process of handling conflict detection according to the candidate arbitration instructions includes: processing the conflict detection level according to the candidate arbitration instructions to obtain the conflict detection result; Based on the conflict detection results and candidate arbitration instructions, the resolution instructions are processed to obtain the resolved instructions; The resolved instructions are subjected to priority arbitration to obtain priority-arbitrated instructions.

5. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, Health satisfaction analysis is performed on the priority arbitration instructions, including: processing the health feature vector and resource parameter vector based on the priority arbitration instructions through health feature parameters; Based on the equipment health feature vector and priority arbitration instructions, the equipment health degradation amount is processed to obtain the predicted health degradation amount; Resource condition satisfaction is analyzed based on the resource parameter vector and priority arbitration instructions to obtain resource satisfaction.

6. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, Based on the equipment health feature vector and resource satisfaction, the deviation verification process is performed, including: generating a historical deviation sequence based on the deviation sequence of the equipment health feature vector and resource parameter vector; Adaptive threshold processing is performed based on historical deviation sequences to generate adaptive threshold factors. The adaptive threshold factor, predicted health degradation, and resource satisfaction are comprehensively verified to obtain the pre-verification pass instruction.

7. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, Steady-state deviation correction accuracy processing is performed on the instantaneous deviation vector and the measured response delay, including: obtaining the steady-state prediction deviation by processing the instantaneous deviation vector through steady-state deviation trend; Compensation and correction are performed based on steady-state prediction deviation and measured response time delay to obtain the compensation and correction amount; The compensation correction amount and the pre-verification are subjected to secondary correction and accuracy processing through instructions to obtain the execution feedback result.

8. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, Based on the execution feedback results, multi-source auditing is performed, including: generating a consistency comparison report by comparing the consistency of multi-source results based on the execution feedback results; Based on the consistency comparison report, perform digital non-repudiation signature processing to obtain the signed record; The signed records are subjected to blockchain auditing to obtain audit logs.

9. The method for remote control and safety verification of new energy equipment according to claim 1, characterized in that, The original control command includes the target device identifier, operation type, operation parameters, source priority level, and request timestamp.

10. A system applied to the remote control and safety verification method for new energy equipment according to any one of claims 1-9, characterized in that, The system includes: The parsing module is used to collect the original control commands of the target device, and to obtain the parsed control commands by parsing, verifying and initially screening the permissions based on the original control commands. The priority arbitration module is used to process multi-source queue candidates based on parsed control instructions to obtain candidate arbitration instructions, and to perform conflict detection arbitration processing based on the candidate arbitration instructions to obtain priority arbitration instructions. The pre-verification module is used to perform health satisfaction analysis on the instructions after priority arbitration, obtain the equipment health feature vector and resource satisfaction, and obtain the pre-verification pass instruction based on the equipment health feature vector and resource satisfaction through the verification deviation verification process. The execution feedback module is used to perform high-precision delay processing based on the pre-verification through instructions to obtain the instantaneous deviation vector and the measured response delay. The instantaneous deviation vector and the measured response delay are then processed for steady-state deviation correction accuracy to obtain the execution feedback result. The audit module is used to obtain audit logs through multi-source audit processing based on the execution feedback results.

Images