Transparent management method and system for assessment and compensation information of new energy power plant

Abstract

The invention provides a new energy power plant assessment and compensation information transparent management method and system, and relates to the technical field of new energy power plant management, and the method comprises the steps: predicting a power generation capability reference value through a deep learning model, and constructing a power generation capability assessment model based on the reference value, and calculating a comprehensive assessment reference value in combination with the real-time power generation data and a preset assessment rule, performing grading assessment on the power plant, and generating a report. And a cause-and-effect analysis model is utilized to judge a power generation loss reason, if the power generation loss reason is caused by power grid dispatching limitation, compensation power data is calculated, and related data is written into the alliance block chain. And the intelligent contract automatically calculates a compensation amount, calculates an operation efficiency index based on historical data, generates a transparent management report, and distributes the transparent management report to users with different permissions through an encryption channel.

Description

Technical Field

[0001] The present invention relates to a new energy power plant management technology, and in particular to a new energy power plant assessment and compensation information transparent management method and system. Background Art

[0002] With the widespread application of new energy power plants, how to effectively evaluate and manage their power generation capacity and operational efficiency has become a key issue. Traditional power plant assessment methods usually rely on static indicators and cannot dynamically reflect the impact of environmental changes on power generation capacity. To solve this problem, a deep learning model is used to analyze meteorological environment and real-time power generation data, which can more accurately predict power generation capacity and conduct comprehensive assessments.

[0003] In the existing technology, the analysis of the causes of power generation loss is often not comprehensive enough, especially the impact of grid dispatch restrictions on power generation is not fully considered. Through the causal analysis model, the causes of power generation loss can be clearly identified, and the corresponding compensation power can be calculated based on this, providing a more transparent and fair management mechanism for power plant operations.

[0004] In addition, storing data in the alliance blockchain network and automatically calculating and distributing compensation information through smart contracts not only improves data security and transparency, but also simplifies the management process. This method helps new energy power plants optimize operational efficiency in a complex power grid environment. Summary of the invention

[0005] The embodiment of the present invention provides a method and system for transparent management of assessment and compensation information of a new energy power plant, which can solve the problems in the prior art.

[0006] According to a first aspect of the embodiments of the present invention,

[0007] Provide a transparent management method for assessment and compensation information of new energy power plants, including:

[0008] Acquire the real-time power generation data, meteorological environment monitoring data and power grid dispatch instruction data of the new energy power plant, use the deep learning model to analyze the meteorological environment monitoring data, predict the power generation capacity benchmark value of the specified time period, build a power generation capacity assessment model based on the power generation capacity benchmark value, and calculate the comprehensive assessment benchmark value of the new energy power plant through power generation efficiency indicators and power quality indicators according to the real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules;

[0009] Based on the comprehensive assessment benchmark value, a hierarchical assessment is conducted on the new energy power plant to generate a power generation capacity assessment report. According to the power generation capacity assessment report and the grid dispatch instruction data, a causal analysis model is used to determine the cause of the power generation loss of the new energy power plant. When it is determined that the power generation loss is caused by grid dispatch restrictions, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation. The real-time power generation data, meteorological environment monitoring data, power generation capacity assessment report and compensation power data are written into the alliance blockchain network;

[0010] A smart contract with hierarchical permissions is deployed in the alliance blockchain network. The smart contract reads the compensation electricity data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network, and generates an operation efficiency analysis result of the new energy power plant. A transparent management report is generated based on the operation efficiency analysis result, the power generation capacity assessment report, the compensation electricity data and the compensation amount, and the transparent management report is distributed to user terminals with different authority levels through an encrypted channel according to the preset data access rights.

[0011] In an optional embodiment,

[0012] Based on the power generation capacity benchmark value, a power generation capacity assessment model is constructed. According to real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules, the comprehensive assessment benchmark value of the new energy power plant is calculated through power generation efficiency indicators and power quality indicators, including:

[0013] Based on the power generation capacity benchmark value, a power generation capacity assessment model including a double-layer time window is constructed, wherein the double-layer time window includes an inner time window and an outer time window, wherein the inner time window weights the short-term power generation power by a nonlinear time decay function to obtain a short-term power generation characteristic, and the outer time window weights the long-term power generation power by an exponential decreasing function to obtain a long-term power generation characteristic, and a power generation efficiency index is calculated based on a weighted combination of the short-term power generation characteristic and the long-term power generation characteristic;

[0014] Acquire real-time power generation data, process the real-time power generation data using a multi-scale analysis method, obtain voltage deviation rate, three-phase voltage imbalance, frequency deviation, frequency change rate, total harmonic distortion rate and harmonic content, and normalize the processing results to construct a power quality feature vector;

[0015] Inputting the power quality feature vector into a residual neural network for feature extraction, and performing attention weighting on the feature extraction result to obtain a power quality evaluation index;

[0016] A hybrid evolutionary optimization algorithm is constructed, the weights of the power generation efficiency index and the power quality assessment index are encoded into real number vectors, a scaling factor is calculated based on population diversity and fitness change rate, the real number vector is adaptively mutated using the scaling factor, a differential evolution operation is combined with a particle swarm velocity update to perform a local search, the individual crowding degree is calculated by non-dominated sorting to construct a Pareto frontier, and an optimal weight vector is selected from the Pareto frontier;

[0017] A three-layer fuzzy rule base is established based on preset assessment rules, the power generation efficiency index and the power quality index are evaluated in layers, the evaluation results are weighted using the optimal weight vector, and confidence correction is performed in combination with Gaussian process regression prediction to obtain a comprehensive assessment benchmark value for the new energy power plant.

[0018] In an optional embodiment,

[0019] A three-layer fuzzy rule base is established based on the preset assessment rules, the power generation efficiency index and the power quality index are evaluated in layers, the evaluation results are weighted by the optimal weight vector, and the confidence correction is performed in combination with Gaussian process regression prediction, and the comprehensive assessment benchmark values ​​of the new energy power plant are obtained, including:

[0020] A three-layer fuzzy rule base is constructed based on the preset assessment rules, wherein the first-layer rule base contains input language variables and output language variables of the power generation efficiency index, the input language variables include five levels: ultra-low, low, medium, high and extremely high, and a Gaussian membership function is constructed based on the value range of the power generation efficiency index, and the output language variables correspond to scoring levels with different scores;

[0021] The input language variables of the second-layer rule base include five levels: unqualified, basically qualified, qualified, good and excellent. A triangular membership function is constructed based on the value interval of the power quality index. The vertices and support set of the triangular membership function are dynamically optimized through a genetic algorithm. The output language variables correspond to five scoring levels and are defuzzified using the maximum membership method.

[0022] The power generation efficiency index is input into the first layer rule base for fuzzy reasoning to obtain the first layer evaluation value, and the power quality index is input into the second layer rule base for fuzzy reasoning to obtain the second layer evaluation value;

[0023] The third-layer rule base receives the first-layer evaluation value and the second-layer evaluation value as input, and uses a predetermined optimal weight vector to perform a weighted combination of the first-layer evaluation value and the second-layer evaluation value to obtain an initial comprehensive evaluation value;

[0024] The power generation efficiency index and the power quality index form an input feature vector, the initial comprehensive evaluation value is used as an output label, a deep regression model including multiple hidden layers is constructed, an attention calculation unit is set in each hidden layer of the deep regression model to extract the correlation feature of the input feature vector, and the deep regression model is trained based on the correlation feature and the initial comprehensive evaluation value;

[0025] The new power generation efficiency index and power quality index are input into the trained deep regression model to obtain the regression prediction value, the prediction error between the regression prediction value and the actual value is calculated, the confidence weight is calculated based on the prediction error, the initial comprehensive evaluation value and the regression prediction value are weighted and fused through the confidence weight to obtain the comprehensive assessment benchmark value of the new energy power plant.

[0026] In an optional embodiment,

[0027] According to the power generation capacity assessment report and the grid dispatch instruction data, a causal analysis model is used to determine the cause of the power generation loss of the new energy power plant. When it is determined that the power generation loss is caused by grid dispatch restrictions, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation, including:

[0028] Obtaining system parameters and grid dispatch instruction data in the power generation capacity assessment report, aligning them according to timestamps, building a causal analysis model, and using the causal analysis model to extract features from the system parameters and grid dispatch instruction data to obtain a correlation feature sequence;

[0029] A dynamic Bayesian network is constructed based on the associated feature sequence, the power generation equipment state, the power grid operation state and the dispatch instruction information are used as network nodes, a state transition matrix is ​​established to characterize the dependency relationship between the network nodes, and an iterative reweighted parameter learning algorithm is used to optimize the state transition matrix to obtain the node dependency strength;

[0030] The temporal causal strength is calculated based on the node dependency strength, and the dependency strength is accumulated in a sliding time window to obtain the cumulative causal strength. The temporal causal strength and the cumulative causal strength are merged by using adaptive weights to obtain the comprehensive causal strength. A piecewise decision function is constructed according to the comprehensive causal strength to determine whether the power generation loss is caused by grid dispatch restrictions;

[0031] When it is determined that the power generation loss is caused by the grid dispatch restriction, a power generation equipment state vector is constructed based on the associated characteristic sequence, the power generation equipment state vector and the dependency strength are input into a state predictor based on a high-order Taylor series, a power generation capacity benchmark model is established according to an output result of the state predictor, a correction coefficient is calculated using a recursive residual compensation algorithm based on the dependency strength, and the correction coefficient is multiplied by the output value of the power generation capacity benchmark model to obtain a theoretical power generation;

[0032] A compensation optimization objective function is constructed according to the deviation between the theoretical power generation and the actual power generation, a dynamic relaxation factor is introduced into the compensation optimization objective function to adjust the compensation intensity, an improved gradient projection algorithm is used to solve the compensation coefficient, and the product of the compensation coefficient and the lost electricity is determined as the final compensation electricity data of the new energy power plant.

[0033] In an optional embodiment,

[0034] Inputting the state vector of the power generation equipment and the dependency strength into a state predictor based on a high-order Taylor series, establishing a power generation capacity benchmark model according to the output result of the state predictor, calculating a correction coefficient using a recursive residual compensation algorithm based on the dependency strength, and multiplying the correction coefficient by the output value of the power generation capacity benchmark model to obtain a theoretical power generation includes:

[0035] Decompose the state vector and the dependency strength of the power generation equipment by wavelet transform to obtain the first characteristic component and the second characteristic component, determine the Taylor series expansion order corresponding to each characteristic component based on the minimum description length criterion, and perform Taylor series expansion on the first characteristic component and the state mapping function of the first characteristic component to obtain the first characteristic prediction sequence and the second characteristic prediction sequence;

[0036] Projecting the dependency intensity matrix onto a Riemann manifold space to obtain a geometric eigenvector, constructing a dynamic coupling weight matrix based on the geometric eigenvector, and using the dynamic coupling weight matrix to perform a weighted combination of the first feature prediction sequence and the second feature prediction sequence to obtain a state prediction vector;

[0037] Input the state prediction vector into a pre-built deep probability model to extract probability distribution features, perform multi-head time series attention calculation on the probability distribution features to obtain time series correlation features, build a power generation capacity benchmark model based on the time series correlation features, and calculate the benchmark power generation through the power generation capacity benchmark model;

[0038] Projecting the difference sequence between the reference power generation and the actual power generation into the orthogonal wavelet basis space to obtain multi-level residual components, constructing residual compensation coefficients based on the characteristic decomposition results of the dependency intensity, and using the residual compensation coefficients to dynamically compensate the multi-level residual components;

[0039] Constructing a graph attention compensation network, inputting the compensated multi-level residual components into the graph attention compensation network, obtaining a dynamic correction factor sequence through recursive operation, and establishing a time series consistency evaluation model at the same time, multiplying the dynamic correction factor sequence with the benchmark power generation to obtain a candidate theoretical power generation, and calculating the distribution deviation between the candidate theoretical power generation and the actual power generation based on the time series consistency evaluation model;

[0040] An adaptive optimizer based on Riemann metric is constructed, and the adaptive optimizer is used to jointly optimize the state prediction vector, the benchmark power generation and the dynamic correction factor sequence, so as to minimize the distribution deviation and obtain the final theoretical power generation.

[0041] In an optional embodiment,

[0042] A smart contract with hierarchical permissions is deployed in the alliance blockchain network. The smart contract reads the compensation power data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power generation conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network. The operation efficiency analysis results of the new energy power plant are generated, including:

[0043] Deploy a hierarchical permission smart contract with management layer, execution layer and query layer in the alliance blockchain network, where the management layer uses asymmetric keys for identity authentication, the execution layer sets the compensation rule state machine, the query layer configures data access rights, and generates a contract deployment identifier;

[0044] Input the contract deployment identifier and the compensation power data into the compensation rule state machine, calculate the compensation benchmark value based on the segmented mapping function, use a deep neural network to predict the compensation adjustment factor, and multiply the compensation benchmark value by the compensation adjustment factor to obtain an initial compensation amount;

[0045] The initial compensation amount and the historical compensation record form a compensation state vector, a temporal attention network is used to extract the temporal features of the compensation state vector, a compensation constraint model is constructed based on the temporal features, and the compensation amount that satisfies the compensation constraint model is written into the alliance blockchain network;

[0046] Read the operation data of the new energy power plant from the alliance blockchain network and perform wavelet decomposition to obtain an operation characteristic sequence, calculate the power generation conversion efficiency index based on the operation characteristic sequence, and calculate the equipment utilization index based on the equipment operation time to obtain an efficiency index set;

[0047] Input the efficiency index set into the graph convolutional neural network to extract spatial correlation features, build a new energy power plant efficiency evaluation model and calculate the operating efficiency numerical sequence;

[0048] The operating efficiency numerical sequence is segmented into sliding time windows, the efficiency change trend in each time window is calculated using the exponential smoothing method, a power plant similarity matrix is ​​constructed based on the efficiency change trend, and the power plants are grouped using a density clustering algorithm;

[0049] The probability distribution of the operating efficiency of each power plant group is calculated, the degree of deviation is identified using the local anomaly factor algorithm, efficiency anomaly features are generated, the efficiency change trend and the efficiency anomaly features are input into the deep neural network to generate the operating efficiency analysis results of the new energy power plant.

[0050] In an optional embodiment,

[0051] Distributing the transparent management report to user terminals of different authority levels through an encrypted channel according to preset data access rights includes:

[0052] Construct a multi-level permission system including system management permission, operation management permission, technical personnel permission, partner permission and public permission, set up an access control matrix based on the multi-level permission system, generate a permission management contract using identity authentication information and timestamp, and deploy the permission management contract to the blockchain network;

[0053] The transparent management report is divided into core data, commercial data and public data according to data sensitivity, and encrypted using differentiated encryption strategies such as group encryption, block encryption and stream encryption to obtain encrypted data;

[0054] Generate a master key, generate a hierarchical subkey sequence for the multi-level authority system based on the master key using a hierarchical key derivation algorithm, establish a unidirectional inheritance relationship between the hierarchical subkey sequences and embed a time constraint condition, and construct a hierarchical key derivation tree;

[0055] Receive an access request sent by a user terminal, verify the identity of the user terminal using dual identity authentication and behavior feature recognition, perform key negotiation with the authenticated user terminal to generate a session key, and establish a two-way authentication encryption channel based on the session key;

[0056] Obtaining the responsibility information, behavior characteristics and operation intention of the user terminal, establishing an access policy model, mapping the access policy model to a rights management contract for rights determination, and obtaining a target key from the hierarchical key derivation tree based on the determination result;

[0057] The encrypted data is decrypted using the target key to obtain the target data, the target data is distributed to the user terminal through the two-way authentication encryption channel, and an access certificate including an identity, an operation record and a timestamp is generated, and the access certificate is written into the blockchain network to form a traceable access chain.

[0058] According to a second aspect of the embodiments of the present invention,

[0059] Provide a transparent management system for assessment and compensation information of new energy power plants, including:

[0060] The first unit is used to obtain real-time power generation data, meteorological environment monitoring data and power grid dispatch instruction data of new energy power plants, use a deep learning model to analyze the meteorological environment monitoring data, predict the power generation capacity benchmark value of a specified period, build a power generation capacity assessment model based on the power generation capacity benchmark value, and calculate the comprehensive assessment benchmark value of the new energy power plant through power generation efficiency indicators and power quality indicators according to real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules;

[0061] The second unit is used to perform a hierarchical assessment of the new energy power plant based on the comprehensive assessment benchmark value, generate a power generation capacity assessment report, and use a causal analysis model to determine the cause of the power generation loss of the new energy power plant based on the power generation capacity assessment report and the power grid dispatch instruction data. When it is determined that the power generation loss is caused by the power grid dispatch restriction, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation, and the real-time power generation data, meteorological environment monitoring data, power generation capacity assessment report and compensation power data are written into the alliance blockchain network;

[0062] The third unit is used to deploy a smart contract with hierarchical permissions in the alliance blockchain network, the smart contract reads the compensation electricity data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network, generates the operation efficiency analysis result of the new energy power plant, generates a transparent management report based on the operation efficiency analysis result, the power generation capacity evaluation report, the compensation electricity data and the compensation amount, and distributes the transparent management report to user terminals with different authority levels through an encrypted channel according to the preset data access rights.

[0063] According to a third aspect of the embodiments of the present invention,

[0064] An electronic device is provided, comprising:

[0065] processor;

[0066] a memory for storing processor-executable instructions;

[0067] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0068] A fourth aspect of the embodiments of the present invention is:

[0069] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0070] In this embodiment, BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a flow chart of a transparent management method for assessment and compensation information of a new energy power plant according to an embodiment of the present invention;

[0072] Figure 2 It is a structural diagram of a transparent management system for assessment and compensation information of a new energy power plant according to an embodiment of the present invention. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0074] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0075] Figure 1 FIG. 1 is a flow chart of a transparent management method for assessment and compensation information of a new energy power plant according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0076] S101. Acquire real-time power generation data, meteorological environment monitoring data and power grid dispatch instruction data of new energy power plants, use deep learning models to analyze the meteorological environment monitoring data, predict the power generation capacity benchmark value of a specified period, build a power generation capacity assessment model based on the power generation capacity benchmark value, and calculate the comprehensive assessment benchmark value of new energy power plants through power generation efficiency indicators and power quality indicators according to real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules;

[0077] S102. Based on the comprehensive assessment benchmark value, a hierarchical assessment is conducted on the new energy power plant, and a power generation capacity assessment report is generated. According to the power generation capacity assessment report and the grid dispatch instruction data, a causal analysis model is used to determine the cause of the power generation loss of the new energy power plant. When it is determined that the power generation loss is caused by grid dispatch restrictions, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation. The real-time power generation data, meteorological environment monitoring data, power generation capacity assessment report and compensation power data are written into the alliance blockchain network;

[0078] S103. A smart contract with hierarchical permissions is deployed in the alliance blockchain network. The smart contract reads the compensation electricity data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network, generates the operation efficiency analysis results of the new energy power plant, generates a transparent management report based on the operation efficiency analysis results, the power generation capacity assessment report, the compensation electricity data and the compensation amount, and distributes the transparent management report to user terminals with different authority levels through an encrypted channel according to the preset data access rights.

[0079] Among them, the real-time power generation data of new energy power plants refers to the operating data such as power generation and output power obtained from new energy power plants at the current time and within the preset time interval, so as to monitor and analyze the operating status of the power plants in real time; meteorological environment monitoring data refers to the data collection of meteorological factors (such as temperature, humidity, wind speed, sunshine intensity, etc.) in the power generation area, which is used to analyze and predict the impact on power generation capacity; power grid dispatching instruction data refers to the operating adjustment instruction data issued by the power grid operator to ensure the balance, safety and stability of the system, including load dispatching and output limitation.

[0080] The power generation capacity benchmark value refers to the theoretical maximum power generation or expected power generation capacity of a new energy power plant under current meteorological conditions, which is used to evaluate power generation efficiency and scheduling potential; the power generation capacity assessment model is a mathematical or statistical model that combines meteorological and power generation capacity benchmark values, which is used to dynamically evaluate and adjust power generation capacity to ensure the optimal operation of new energy power plants under different conditions.

[0081] The consortium blockchain network is a distributed ledger system with multi-party participation and restricted access rights. Each participant shares data but has different permissions to ensure data security and transparency; layered permissions are a management method for hierarchical control of access and operating permissions. Users of different levels have different operating rights on the blockchain; a smart contract is an automatically executed contract based on the blockchain with a conditional judgment function, which is used to automatically calculate the compensation amount and achieve self-execution.

[0082] In an optional implementation, a power generation capacity assessment model is constructed based on the power generation capacity benchmark value, and according to real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules, the comprehensive assessment benchmark value of the new energy power plant is calculated through power generation efficiency indicators and power quality indicators, including:

[0083] Based on the power generation capacity benchmark value, a power generation capacity assessment model including a double-layer time window is constructed, wherein the double-layer time window includes an inner time window and an outer time window, wherein the inner time window weights the short-term power generation power by a nonlinear time decay function to obtain a short-term power generation characteristic, and the outer time window weights the long-term power generation power by an exponential decreasing function to obtain a long-term power generation characteristic, and a power generation efficiency index is calculated based on a weighted combination of the short-term power generation characteristic and the long-term power generation characteristic;

[0084] Acquire real-time power generation data, process the real-time power generation data using a multi-scale analysis method, obtain voltage deviation rate, three-phase voltage imbalance, frequency deviation, frequency change rate, total harmonic distortion rate and harmonic content, and normalize the processing results to construct a power quality feature vector;

[0085] Inputting the power quality feature vector into a residual neural network for feature extraction, and performing attention weighting on the feature extraction result to obtain a power quality evaluation index;

[0086] A hybrid evolutionary optimization algorithm is constructed, the weights of the power generation efficiency index and the power quality assessment index are encoded into real number vectors, a scaling factor is calculated based on population diversity and fitness change rate, the real number vector is adaptively mutated using the scaling factor, a differential evolution operation is combined with a particle swarm velocity update to perform a local search, the individual crowding degree is calculated by non-dominated sorting to construct a Pareto frontier, and an optimal weight vector is selected from the Pareto frontier;

[0087] A three-layer fuzzy rule base is established based on preset assessment rules, the power generation efficiency index and the power quality index are evaluated in layers, the evaluation results are weighted using the optimal weight vector, and confidence correction is performed in combination with Gaussian process regression prediction to obtain a comprehensive assessment benchmark value for the new energy power plant.

[0088] Exemplarily, first, a power generation capacity benchmark is constructed. In order to more accurately reflect the power generation capacity of the power plant, a power generation capacity assessment model with a double-layer time window is adopted. The inner time window focuses on short-term power generation, such as power generation data in the past week, and gives higher weights to recent data, and the weights decrease nonlinearly over time, such as using a decay method similar to an inverse proportional function to highlight the impact of short-term power generation performance. The outer time window focuses on long-term power generation, such as power generation data in the past year, and gives higher weights to recent data, and the weights decrease exponentially over time to reflect the stability of long-term power generation capacity. For example, the negative t power of e (t is the time interval) can be used as the decay function. The short-term and long-term power generation characteristics are weighted and combined, such as a ratio of 7:3, to obtain the final power generation efficiency index. Assuming that the power generation of a power plant in the past week is 100MW, 110MW, 105MW, 95MW, 100MW, 115MW, and 120MW, the short-term power generation characteristic value is obtained after nonlinear time decay weighting. Assuming that the average power generation per week in the past year is 100MW, the long-term power generation characteristic value is obtained after exponentially decreasing weighting. Finally, the short-term power generation characteristic value and the long-term power generation characteristic value are weighted and combined to obtain the power generation efficiency index.

[0089] Next, acquire and process real-time power generation data. Real-time power generation data includes information such as voltage, current, and frequency. A multi-scale analysis method is used to process real-time power generation data. For example, wavelet transform can be used to decompose and reconstruct data to extract features at different time scales. The voltage deviation rate is calculated by analyzing the deviation between the voltage amplitude and the rated voltage. The three-phase voltage imbalance is calculated by analyzing the difference between the three-phase voltages. The frequency deviation is calculated by analyzing the deviation between the frequency and the rated frequency. The frequency change rate is calculated by analyzing the rate of change of the frequency over time. The total harmonic distortion rate and the harmonic content of each order are calculated by analyzing the ratio of the harmonic component to the fundamental component. The indicators such as the voltage deviation rate, the three-phase voltage imbalance, the frequency deviation, the frequency change rate, the total harmonic distortion rate, and the harmonic content are normalized, for example, all indicators are scaled to between 0 and 1 to construct a power quality feature vector. Assuming that the voltage deviation rate at a certain moment is 0.02, the three-phase voltage unbalance is 0.01, the frequency deviation is 0.005, the frequency change rate is 0.001, and the total harmonic distortion rate is 0.05, then after normalization, the power quality characteristic vector [0.02, 0.01, 0.005, 0.001, 0.05] is obtained.

[0090] Then, the residual neural network is used to extract the power quality features. The power quality feature vector is input into the residual neural network for feature extraction. The residual neural network can effectively solve the gradient vanishing problem in the deep network training process through cross-layer connection, so as to extract deeper features. The feature extraction results are weighted by attention. The attention mechanism can assign different weights according to the importance of the features, thereby improving the accuracy of the model. Finally, the power quality evaluation index is obtained.

[0091] After that, the weights of the power generation efficiency index and the power quality assessment index are determined. A hybrid evolutionary optimization algorithm is constructed to encode the weights of the power generation efficiency index and the power quality assessment index into real vectors. The scaling factor is calculated based on the population diversity and the fitness change rate. The real vector is adaptively mutated using the scaling factor. The differential evolution operation is combined with the particle swarm velocity update for local search. The Pareto front is constructed by calculating the individual crowding degree through non-dominated sorting. The optimal weight vector is selected from the Pareto front. For example, assuming that the weight of the power generation efficiency index is 0.8 and the weight of the power quality assessment index is 0.2, the optimal weight vector is [0.8, 0.2].

[0092] Finally, the comprehensive assessment benchmark value of the new energy power plant is calculated. A three-layer fuzzy rule base is established based on the preset assessment rules to perform hierarchical assessments on the power generation efficiency index and the power quality index. For example, according to the different value ranges of the power generation efficiency index and the power quality index, they can be divided into four levels: excellent, good, qualified, and unqualified. The evaluation results are weighted using the optimal weight vector. Combined with Gaussian process regression prediction, confidence correction is performed to obtain the comprehensive assessment benchmark value of the new energy power plant. For example, assuming that the evaluation result of the power generation efficiency index is excellent and the evaluation result of the power quality index is good, then after weighting and confidence correction, the comprehensive assessment benchmark value of the new energy power plant is 95 points.

[0093] In this embodiment, through the double-layer time window and multi-scale analysis method, the short-term and long-term power generation performance of the power plant and various aspects of power quality can be considered more comprehensively, thereby improving the accuracy of the evaluation model. The hybrid evolutionary optimization algorithm can effectively search for the optimal weight vector, thereby improving the reliability of the evaluation results. The fuzzy rule base and confidence correction can effectively reduce the influence of human factors, thereby improving the objectivity of the evaluation results.

[0094] In an optional implementation, a three-layer fuzzy rule base is established based on preset assessment rules, the power generation efficiency index and the power quality index are evaluated in layers, the evaluation results are weighted by the optimal weight vector, and the confidence correction is performed in combination with Gaussian process regression prediction, and the comprehensive assessment benchmark value of the new energy power plant is obtained, including:

[0095] A three-layer fuzzy rule base is constructed based on the preset assessment rules, wherein the first-layer rule base contains input language variables and output language variables of the power generation efficiency index, the input language variables include five levels: ultra-low, low, medium, high and extremely high, and a Gaussian membership function is constructed based on the value range of the power generation efficiency index, and the output language variables correspond to scoring levels with different scores;

[0096] The input language variables of the second-layer rule base include five levels: unqualified, basically qualified, qualified, good and excellent. A triangular membership function is constructed based on the value interval of the power quality index. The vertices and support set of the triangular membership function are dynamically optimized through a genetic algorithm. The output language variables correspond to five scoring levels and are defuzzified using the maximum membership method.

[0097] The power generation efficiency index is input into the first layer rule base for fuzzy reasoning to obtain the first layer evaluation value, and the power quality index is input into the second layer rule base for fuzzy reasoning to obtain the second layer evaluation value;

[0098] The third-layer rule base receives the first-layer evaluation value and the second-layer evaluation value as input, and uses a predetermined optimal weight vector to perform a weighted combination of the first-layer evaluation value and the second-layer evaluation value to obtain an initial comprehensive evaluation value;

[0099] The power generation efficiency index and the power quality index form an input feature vector, the initial comprehensive evaluation value is used as an output label, a deep regression model including multiple hidden layers is constructed, an attention calculation unit is set in each hidden layer of the deep regression model to extract the correlation feature of the input feature vector, and the deep regression model is trained based on the correlation feature and the initial comprehensive evaluation value;

[0100] The new power generation efficiency index and power quality index are input into the trained deep regression model to obtain the regression prediction value, the prediction error between the regression prediction value and the actual value is calculated, the confidence weight is calculated based on the prediction error, the initial comprehensive evaluation value and the regression prediction value are weighted and fused through the confidence weight to obtain the comprehensive assessment benchmark value of the new energy power plant.

[0101] Exemplarily, first, a three-layer fuzzy rule base is established. The first layer of the rule base is used to evaluate the power generation efficiency index. The value range of the power generation efficiency index is divided into five levels: ultra-low, lower, medium, higher and extremely high. For example, assuming that the value range of the power generation efficiency is 0% to 100%, "ultra-low" can be defined as 0% to 20%, "low" can be defined as 20% to 40%, and so on. A Gaussian membership function is constructed for each level. For example, for the "medium" level, the central value can be set to 60% and the standard deviation can be set to 10%. The output language variables correspond to scoring levels with different scores, for example, "ultra-low" corresponds to 1 point, "low" corresponds to 3 points, "medium" corresponds to 5 points, "high" corresponds to 7 points, and "extremely high" corresponds to 9 points.

[0102] The second-level rule base is used to evaluate power quality indicators. The input language variables are also divided into five levels: unqualified, basically qualified, qualified, good and excellent. The difference is that the triangular membership function is used here. In order to improve the accuracy of the evaluation, the vertices and support sets of the triangular membership function are dynamically optimized by genetic algorithm. For example, the initial triangular membership function vertex of the "qualified" level is 70, and the left and right support sets are 60 and 80 respectively. After optimization by genetic algorithm, the vertex may become 72, and the support set may become 62 and 82. The output language variables also correspond to five scoring levels, for example, "unqualified" corresponds to 1 point, "basic qualified" corresponds to 3 points, "qualified" corresponds to 5 points, "good" corresponds to 7 points, and "excellent" corresponds to 9 points. The maximum membership method is used for defuzzification. For example, if the membership of a power quality indicator is 0.8 in the "qualified" level and 0.2 in the "good" level, it is finally judged as "qualified".

[0103] The third layer rule base receives the evaluation values ​​from the first two layers of rule bases as input. An optimal weight vector is determined in advance, for example, the weight of the power generation efficiency index is 0.6, and the weight of the power quality index is 0.4. These two evaluation values ​​are multiplied by the corresponding weights and then added to obtain the initial comprehensive evaluation value. For example, if the evaluation value of the power generation efficiency index is 7 points and the evaluation value of the power quality index is 5 points, then the initial comprehensive evaluation value is 7*0.6+5*0.4=6.2 points.

[0104] Next, build a deep regression model. The power generation efficiency index and the power quality index form the input feature vector, and the initial comprehensive evaluation value is used as the output label. The model contains multiple hidden layers, and an attention calculation unit is set in each hidden layer to extract the correlation characteristics of the input feature vector. For example, if the power generation efficiency index and the power quality index are highly correlated, the model will pay more attention to this correlation. Use historical data to train the deep regression model. For example, collect the power generation efficiency index, power quality index and the corresponding initial comprehensive evaluation value for each day in the past year as a training data set.

[0105] Input the new power generation efficiency index and power quality index into the trained deep regression model to obtain the regression prediction value. Calculate the prediction error between the regression prediction value and the actual value (i.e., the initial comprehensive evaluation value). Calculate the confidence weight based on the prediction error. For example, if the prediction error is very small, the confidence weight is high; otherwise, the confidence weight is low. The initial comprehensive evaluation value and the regression prediction value are weighted and fused by the confidence weight to obtain the final comprehensive assessment benchmark value of the new energy power plant. For example, assuming that the initial comprehensive assessment value is 6.2 points, the regression prediction value is 6.5 points, and the confidence weight is 0.9, the final comprehensive assessment benchmark value is 6.2*(1-0.9)+6.5*0.9=6.47 points.

[0106] In this embodiment, by combining the fuzzy rule base and the deep learning model, the advantages of expert experience and data-driven are fully considered, which can more accurately reflect the actual operation of the power plant, thereby improving the accuracy of the assessment results. By presetting the assessment rules and dynamically optimizing the membership function, the interference of human factors is reduced, making the assessment process more objective and fair. The deep regression model can continuously learn and adjust according to new data, so as to adapt to the assessment needs of different types of new energy power plants and improve the adaptability of the assessment method.

[0107] In an optional implementation, based on the power generation capacity assessment report and the grid dispatch instruction data, a causal analysis model is used to determine the cause of the power generation loss of the new energy power plant. When it is determined that the power generation loss is caused by grid dispatch restrictions, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation, including:

[0108] Obtaining system parameters and grid dispatch instruction data in the power generation capacity assessment report, aligning them according to timestamps, building a causal analysis model, and using the causal analysis model to extract features from the system parameters and grid dispatch instruction data to obtain a correlation feature sequence;

[0109] A dynamic Bayesian network is constructed based on the associated feature sequence, the power generation equipment state, the power grid operation state and the dispatch instruction information are used as network nodes, a state transition matrix is ​​established to characterize the dependency relationship between the network nodes, and an iterative reweighted parameter learning algorithm is used to optimize the state transition matrix to obtain the node dependency strength;

[0110] The temporal causal strength is calculated based on the node dependency strength, and the dependency strength is accumulated in a sliding time window to obtain the cumulative causal strength. The temporal causal strength and the cumulative causal strength are merged by using adaptive weights to obtain the comprehensive causal strength. A piecewise decision function is constructed according to the comprehensive causal strength to determine whether the power generation loss is caused by grid dispatch restrictions;

[0111] When it is determined that the power generation loss is caused by the grid dispatch restriction, a power generation equipment state vector is constructed based on the associated characteristic sequence, the power generation equipment state vector and the dependency strength are input into a state predictor based on a high-order Taylor series, a power generation capacity benchmark model is established according to an output result of the state predictor, a correction coefficient is calculated using a recursive residual compensation algorithm based on the dependency strength, and the correction coefficient is multiplied by the output value of the power generation capacity benchmark model to obtain a theoretical power generation;

[0112] A compensation optimization objective function is constructed according to the deviation between the theoretical power generation and the actual power generation, a dynamic relaxation factor is introduced into the compensation optimization objective function to adjust the compensation intensity, an improved gradient projection algorithm is used to solve the compensation coefficient, and the product of the compensation coefficient and the lost electricity is determined as the final compensation electricity data of the new energy power plant.

[0113] Exemplarily, first, extract key system parameters from the power generation capacity assessment report, such as ambient temperature, wind speed, light intensity, equipment operating status, etc. At the same time, extract information such as dispatch instruction type, instruction execution time, target power generation, etc. from the power grid dispatch instruction data. Then, align the extracted system parameters and dispatch instruction data according to the timestamp to ensure the consistency of the data in time. For example, align the data at intervals of 1 minute, and the missing data can be supplemented using linear interpolation. Finally, normalize the aligned data and convert it into a unified numerical range for subsequent model training and analysis.

[0114] Construct a causal analysis model and extract associated feature sequences. Use the preprocessed data to construct a causal analysis model. Here, a causal analysis model based on deep learning is used, such as a convolutional neural network (CNN) or a recurrent neural network (RNN), to extract features from system parameters and grid dispatch instruction data. Specifically, the aligned data is used as the input of the model, and the associated feature sequences related to power generation loss are extracted through model training and learning. For example, the model can learn features such as a sudden drop in wind speed and frequent changes in dispatch instructions.

[0115] Construct a dynamic Bayesian network and calculate the node dependency strength. Construct a dynamic Bayesian network based on the extracted correlation feature sequence. The power generation equipment status, power grid operation status and dispatch instruction information are used as network nodes. The dependency relationship between nodes is characterized by the state transition matrix. For example, if the power generation equipment status changes from "normal" to "fault", the corresponding state transition matrix element value will increase. In order to optimize the state transition matrix, an iterative reweighted parameter learning algorithm is used to continuously adjust the values ​​of the matrix elements, and finally the dependency strength between nodes is obtained. The larger the dependency strength value, the stronger the correlation between nodes.

[0116] Calculate the comprehensive causal strength and determine the cause of power generation loss. Calculate the temporal causal strength based on the node dependency strength. The temporal causal strength reflects the causal relationship between nodes in the time dimension. Accumulate the temporal causal strength in the sliding time window to obtain the cumulative causal strength, which reflects the cumulative impact between nodes over a period of time. Adaptive weights are used to fuse the temporal causal strength and the cumulative causal strength to obtain the comprehensive causal strength. According to the comprehensive causal strength, a piecewise decision function is constructed. If the comprehensive causal strength exceeds the preset threshold, it is determined that the power generation loss is caused by the grid dispatch restriction. For example, if the comprehensive causal strength reaches 0.8 and the preset threshold is 0.7, it is determined that the power generation loss is caused by the grid dispatch restriction.

[0117] Calculate theoretical power generation. When it is determined that the power generation loss is caused by grid dispatch restrictions, construct the power generation equipment state vector based on the associated feature sequence. Input the power generation equipment state vector and the dependency strength into the state predictor based on the high-order Taylor series. According to the output of the state predictor, establish a power generation capacity benchmark model. Calculate the correction coefficient using a recursive residual compensation algorithm based on dependency strength. Multiply the correction coefficient by the output value of the power generation capacity benchmark model to obtain the theoretical power generation. For example, if the baseline model output is 100MW and the correction coefficient is 0.9, the theoretical power generation is 90MW.

[0118] Calculate the compensation power data of the new energy power plant. Construct the compensation optimization objective function based on the deviation between the theoretical power generation and the actual power generation. Introduce the dynamic relaxation factor in the compensation optimization objective function to adjust the compensation intensity. Use the improved gradient projection algorithm to solve the compensation coefficient. The product of the compensation coefficient and the loss power is determined as the final compensation power data of the new energy power plant. For example, if the loss power is 10MW and the compensation coefficient is 0.8, the compensation power is 8MW.

[0119] In this embodiment, by combining the causal analysis model and the power generation capacity assessment model, the cause of the power generation loss can be judged more accurately and the theoretical power generation can be calculated, thereby improving the calculation accuracy of the compensation power and avoiding the errors caused by the traditional method relying on experience judgment. The use of automated algorithms for data processing, model training and compensation power calculation reduces manual intervention, improves work efficiency and reduces the possibility of human errors. The use of technologies such as adaptive weights and dynamic relaxation factors can automatically adjust parameters according to actual conditions, adapt to different power grid environments and operating conditions, and improve the robustness and practicality of the method.

[0120] In an optional implementation, the state vector of the power generation equipment and the dependency strength are input into a state predictor based on a high-order Taylor series, a power generation capacity benchmark model is established according to the output result of the state predictor, a correction coefficient is calculated using a recursive residual compensation algorithm based on the dependency strength, and the correction coefficient is multiplied by the output value of the power generation capacity benchmark model to obtain the theoretical power generation, which includes:

[0121] Decompose the state vector and the dependency strength of the power generation equipment by wavelet transform to obtain the first characteristic component and the second characteristic component, determine the Taylor series expansion order corresponding to each characteristic component based on the minimum description length criterion, and perform Taylor series expansion on the first characteristic component and the state mapping function of the first characteristic component to obtain the first characteristic prediction sequence and the second characteristic prediction sequence;

[0122] Projecting the dependency intensity matrix onto a Riemann manifold space to obtain a geometric eigenvector, constructing a dynamic coupling weight matrix based on the geometric eigenvector, and using the dynamic coupling weight matrix to perform a weighted combination of the first feature prediction sequence and the second feature prediction sequence to obtain a state prediction vector;

[0123] Input the state prediction vector into a pre-built deep probability model to extract probability distribution features, perform multi-head time series attention calculation on the probability distribution features to obtain time series correlation features, build a power generation capacity benchmark model based on the time series correlation features, and calculate the benchmark power generation through the power generation capacity benchmark model;

[0124] Projecting the difference sequence between the reference power generation and the actual power generation into the orthogonal wavelet basis space to obtain multi-level residual components, constructing residual compensation coefficients based on the characteristic decomposition results of the dependency intensity, and using the residual compensation coefficients to dynamically compensate the multi-level residual components;

[0125] Constructing a graph attention compensation network, inputting the compensated multi-level residual components into the graph attention compensation network, obtaining a dynamic correction factor sequence through recursive operation, and establishing a time series consistency evaluation model at the same time, multiplying the dynamic correction factor sequence with the benchmark power generation to obtain a candidate theoretical power generation, and calculating the distribution deviation between the candidate theoretical power generation and the actual power generation based on the time series consistency evaluation model;

[0126] An adaptive optimizer based on Riemann metric is constructed, and the adaptive optimizer is used to jointly optimize the state prediction vector, the benchmark power generation and the dynamic correction factor sequence, so as to minimize the distribution deviation and obtain the final theoretical power generation.

[0127] Exemplarily, first, multiple state parameters of the power generation equipment are collected, such as speed, temperature, pressure, etc., to form a state vector. At the same time, dependent intensity inputs affecting power generation are obtained, such as weather conditions (temperature, wind speed, light intensity, etc.), load demand, etc.

[0128] Then, the collected state vector and dependency strength input are respectively subjected to wavelet transform. Wavelet transform can decompose the signal into components of different frequencies. Here we decompose and obtain two characteristic components: the first characteristic component mainly reflects the low-frequency trend of the signal, and the second characteristic component mainly reflects the high-frequency details of the signal.

[0129] Next, it is necessary to determine the order of Taylor series expansion for each eigencomponent. The higher the order, the higher the fitting accuracy, but the higher the computational complexity. Here, the minimum description length criterion is used to select the optimal order, which balances the complexity of the model and the fitting accuracy. Specifically, Taylor expansions of different orders are calculated, and the order that minimizes the description length is selected as the optimal order. For example, for the first eigencomponent, the 2nd, 3rd, and 4th order Taylor series expansions are used respectively, and the respective description lengths are calculated, and the order with the smallest description length is selected. Assuming that the description length of the 3rd order Taylor series expansion is the smallest, the 3rd order is selected as the Taylor series expansion order for the first eigencomponent. The same operation is performed on the second eigencomponent.

[0130] After determining the order, the state mapping function of the first eigencomponent and the second eigencomponent is subjected to Taylor series expansion to obtain the first eigenprediction sequence and the second eigenprediction sequence. In order to fuse the two eigenprediction sequences, the dependency intensity matrix is ​​projected into the Riemann manifold space to obtain a geometric eigenvector. A dynamic coupling weight matrix is ​​constructed based on the geometric eigenvector. For example, a weight matrix can be constructed based on the distance between the geometric eigenvectors, and the closer the distance, the greater the weight. Then, the first eigenprediction sequence and the second eigenprediction sequence are weightedly combined using the dynamic coupling weight matrix to obtain a state prediction vector.

[0131] The obtained state prediction vector is input into a pre-built deep probability model, such as a variational autoencoder or a generative adversarial network, to extract probability distribution features. Then, a multi-head temporal attention calculation is performed on the extracted probability distribution features to capture the correlation between different time steps and obtain temporal correlation features. A power generation capacity benchmark model is constructed based on the temporal correlation features, such as a recurrent neural network or a Transformer model. The benchmark power generation is calculated using this benchmark model.

[0132] The difference sequence between the reference power generation and the actual power generation is projected into the orthogonal wavelet basis space to obtain multi-level residual components, which can decompose the residual into components of different frequencies. The residual compensation coefficient is constructed based on the eigendecomposition result of the dependency strength. For example, the dependency strength can be subjected to principal component analysis to obtain the main eigenvectors, and then the compensation coefficient is constructed based on these eigenvectors. The residual compensation coefficient is used to dynamically compensate the multi-level residual components, for example, by multiplying the compensation coefficient with the residual component.

[0133] A graph attention compensation network is constructed, and the compensated multi-level residual components are input into the network. The network obtains a dynamic correction factor sequence through recursive operation. At the same time, a temporal consistency evaluation model is established, such as calculating the distribution difference between the candidate theoretical power generation and the actual power generation, such as KL divergence or Wasserstein distance. The dynamic correction factor sequence is multiplied by the benchmark power generation to obtain the candidate theoretical power generation.

[0134] Finally, an adaptive optimizer based on the Riemann metric is constructed, such as the Adam optimizer, which is used to jointly optimize the state prediction vector, the benchmark power generation and the dynamic correction factor sequence to minimize the distribution deviation between the candidate theoretical power generation and the actual power generation to obtain the final theoretical power generation.

[0135] For example, it is assumed that the state vector contains two parameters, temperature and pressure, and the dependence intensity is wind speed. The temperature collected at a certain moment is 30°C, the pressure is 10MPa, and the wind speed is 5m / s. After wavelet transform, the first characteristic component is 25 and the second characteristic component is 5. Assume that the Taylor series expansion order determined by the minimum description length criterion is 2. After a series of calculations, the benchmark power generation is 100MW and the actual power generation is 110MW. The residual is 10MW. After a series of compensation and optimization, the theoretical power generation is finally 108MW.

[0136] In this embodiment, by combining wavelet transform, Taylor series expansion, Riemann manifold projection, deep probability model, multi-head temporal attention mechanism, graph attention compensation network and other technologies, the complex relationship between the state of power generation equipment and the dependence intensity can be captured more accurately, thereby improving the accuracy of power generation prediction. The use of strategies such as dynamic coupling weights, residual compensation, and adaptive optimization can effectively deal with data noise and fluctuations and enhance the robustness and stability of the model. By selecting the Taylor series expansion order based on the minimum description length criterion, the complexity of the model can be effectively controlled, the computational cost can be reduced, and the computational efficiency can be improved.

[0137] In an optional implementation, a smart contract with hierarchical permissions is deployed in the alliance blockchain network, the smart contract reads the compensation power data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power generation conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network, and generates the operating efficiency analysis results of the new energy power plant, including:

[0138] Deploy a hierarchical permission smart contract with management layer, execution layer and query layer in the alliance blockchain network, where the management layer uses asymmetric keys for identity authentication, the execution layer sets the compensation rule state machine, the query layer configures data access rights, and generates a contract deployment identifier;

[0139] Input the contract deployment identifier and the compensation power data into the compensation rule state machine, calculate the compensation benchmark value based on the segmented mapping function, use a deep neural network to predict the compensation adjustment factor, and multiply the compensation benchmark value by the compensation adjustment factor to obtain an initial compensation amount;

[0140] The initial compensation amount and the historical compensation record form a compensation state vector, a temporal attention network is used to extract the temporal features of the compensation state vector, a compensation constraint model is constructed based on the temporal features, and the compensation amount that satisfies the compensation constraint model is written into the alliance blockchain network;

[0141] Read the operation data of the new energy power plant from the alliance blockchain network and perform wavelet decomposition to obtain an operation characteristic sequence, calculate the power generation conversion efficiency index based on the operation characteristic sequence, and calculate the equipment utilization index based on the equipment operation time to obtain an efficiency index set;

[0142] Input the efficiency index set into the graph convolutional neural network to extract spatial correlation features, build a new energy power plant efficiency evaluation model and calculate the operating efficiency numerical sequence;

[0143] The operating efficiency numerical sequence is segmented into sliding time windows, the efficiency change trend in each time window is calculated using the exponential smoothing method, a power plant similarity matrix is ​​constructed based on the efficiency change trend, and the power plants are grouped using a density clustering algorithm;

[0144] The probability distribution of the operating efficiency of each power plant group is calculated, the degree of deviation is identified using the local anomaly factor algorithm, efficiency anomaly features are generated, the efficiency change trend and the efficiency anomaly features are input into the deep neural network to generate the operating efficiency analysis results of the new energy power plant.

[0145] In this embodiment, first, a smart contract based on the alliance chain is constructed to realize hierarchical permission management. The contract includes a management layer, an execution layer, and a query layer. The management layer is responsible for the deployment and update of the contract, and uses an asymmetric encryption algorithm for identity authentication, such as the RSA algorithm. Only administrators with private keys can modify the contract. Assume that the administrator's public key is A and the private key is B. The execution layer is responsible for processing the state machine of the compensation rule. The state machine contains four states: pending review, approved, rejected, and completed, and performs corresponding operations according to different states. The query layer allows external users to access some data, such as historical compensation records and power plant efficiency indicators, but cannot modify the contract content. After the contract is deployed, a unique identifier is generated, such as a hash value C.

[0146] Next, calculate the compensation amount. Input the contract deployment identifier C and the compensation power data into the compensation rule state machine. The calculation of the compensation benchmark value adopts a segmented mapping function. For example, if the compensation power is less than 1000 kWh, the compensation benchmark value is 0.5 yuan / kWh; if the compensation power is between 1000 kWh and 5000 kWh, the compensation benchmark value is 0.4 yuan / kWh; if the compensation power is greater than 5000 kWh, the compensation benchmark value is 0.3 yuan / kWh. Assuming that the compensation power of a power plant is 2000 kWh, the compensation benchmark value is 0.4 yuan / kWh. The compensation adjustment factor is predicted by a deep neural network, such as a three-layer neural network, with inputs such as historical electricity market prices and weather data, and outputs the compensation adjustment factor. Assume that the predicted compensation adjustment factor is 1.2. Multiply the compensation benchmark value by the compensation adjustment factor to obtain the initial compensation amount, that is, 0.4 yuan / kWh*2000 kWh*1.2=960 yuan.

[0147] Then, the initial compensation amount and the historical compensation record are combined into a compensation state vector. For example, the compensation amount of the past 12 months and the initial compensation amount of this time are combined into a 13-dimensional vector. The temporal attention network is used to extract the temporal features of the compensation state vector, such as extracting the weight of the compensation amount of each month. A compensation constraint model is constructed based on the extracted temporal features, such as limiting the compensation amount to no more than twice the compensation amount of the previous month. If the constraint model is satisfied, the compensation amount is written into the alliance blockchain; if not, the modified compensation amount is returned.

[0148] Next, analyze the operating efficiency of the new energy power plant. Read the operating data of the power plant from the alliance blockchain, such as power generation, equipment operating time, etc. Perform wavelet decomposition on the operating data to obtain the operating characteristic sequence, such as decomposing the power generation into subsequences of different frequencies. Calculate the power generation conversion efficiency index based on the operating characteristic sequence, such as dividing the power generation by the amount of energy consumed. At the same time, calculate the equipment utilization index based on the equipment operating time, such as dividing the actual equipment operating time by the planned operating time. Integrate the calculated indicators into an efficiency indicator set. Assume that the power generation conversion efficiency of a power plant is 80% and the equipment utilization rate is 90%.

[0149] After that, the efficiency index set is input into the graph convolutional neural network to extract spatial correlation features, such as considering the geographical location relationship between different power plants. A new energy power plant efficiency evaluation model is constructed and the operating efficiency numerical sequence is calculated, such as the weighted sum of power generation conversion efficiency and equipment utilization to obtain a comprehensive efficiency index and generate a time series.

[0150] Next, analyze the numerical sequence of operational efficiency. Use a sliding time window to segment the time series, for example, one month as a time window. Use the exponential smoothing method to calculate the efficiency change trend for the efficiency values ​​in each time window, for example, calculate the efficiency growth rate for each month. Construct a power plant similarity matrix based on the efficiency change trend, for example, calculate the difference in efficiency growth rates between different power plants. Use a density clustering algorithm to complete the grouping of power plants, for example, group power plants with similar efficiency change trends.

[0151] Finally, the operating efficiency of each power plant group is evaluated. The operating efficiency probability distribution of each group is calculated, such as the mean and standard deviation of the efficiency values ​​of each group of power plants. The local anomaly factor algorithm is used to identify the degree of deviation and generate efficiency anomaly features, such as identifying power plants whose efficiency values ​​deviate significantly from the mean. The efficiency change trend and efficiency anomaly features are input into the deep neural network to generate the operating efficiency analysis results of the new energy power plant, such as predicting the efficiency change trend in the future and giving improvement suggestions.

[0152] In this embodiment, compensation is automatically calculated and issued through smart contracts, human intervention is reduced, efficiency is improved, and the fairness and rationality of compensation are ensured by using the time-series attention network and compensation constraint model. By analyzing the operating data of the power plant, calculating efficiency indicators, identifying efficiency anomalies, and giving improvement suggestions, the power plant is helped to optimize operations, improve power generation efficiency and equipment utilization. By building a platform based on the alliance chain, data sharing and collaborative analysis are realized, the healthy development of the new energy industry is promoted, and energy transformation is promoted.

[0153] In an optional implementation, distributing the transparent management report to user terminals of different authority levels through an encrypted channel according to preset data access rights includes:

[0154] Construct a multi-level permission system including system management permission, operation management permission, technical personnel permission, partner permission and public permission, set up an access control matrix based on the multi-level permission system, generate a permission management contract using identity authentication information and timestamp, and deploy the permission management contract to the blockchain network;

[0155] The transparent management report is divided into core data, commercial data and public data according to data sensitivity, and encrypted using differentiated encryption strategies such as group encryption, block encryption and stream encryption to obtain encrypted data;

[0156] Generate a master key, generate a hierarchical subkey sequence for the multi-level authority system based on the master key using a hierarchical key derivation algorithm, establish a unidirectional inheritance relationship between the hierarchical subkey sequences and embed a time constraint condition, and construct a hierarchical key derivation tree;

[0157] Receive an access request sent by a user terminal, verify the identity of the user terminal using dual identity authentication and behavior feature recognition, perform key negotiation with the authenticated user terminal to generate a session key, and establish a two-way authentication encryption channel based on the session key;

[0158] Obtaining the responsibility information, behavior characteristics and operation intention of the user terminal, establishing an access policy model, mapping the access policy model to a rights management contract for rights determination, and obtaining a target key from the hierarchical key derivation tree based on the determination result;

[0159] The encrypted data is decrypted using the target key to obtain the target data, the target data is distributed to the user terminal through the two-way authentication encryption channel, and an access certificate including an identity, an operation record and a timestamp is generated, and the access certificate is written into the blockchain network to form a traceable access chain.

[0160] Exemplarily, first, a multi-level permission system and access control mechanism are constructed. The system presets five permission levels: system management permission, operation management permission, technical personnel permission, partner permission and public permission. The access control matrix specifies the access rights of different permission levels to different data types. For example, system management permission can access all data types, while public permission can only access public data. When a user logs in to the system, the system generates a permission management contract based on his identity information and the current timestamp. The contract contains information such as the user's identity, permission level and validity period. The generated permission management contract will be deployed to the blockchain network to ensure the immutability and traceability of permission information. For example, user A has operation management permission, and his permission management contract records his identity ID, the permission level is "operation management", and it is valid until December 31, 2024.

[0161] Next, the transparent management report is encrypted differentially. According to the sensitivity of the data, the report is divided into core data, business data and public data. Different encryption strategies are used for different data types: core data uses a combination of group encryption and stream encryption, business data uses block encryption, and public data can be unencrypted or use a lightweight encryption algorithm. For example, the core data "Financial Statement of Project X" is first encrypted in groups and then stream encrypted to generate ciphertext; the business data "Market Research Report" is encrypted using the AES block encryption algorithm; the public data "Company Announcement" is not encrypted.

[0162] Then, a hierarchical key derivation tree is constructed. The system generates a master key, and based on the master key, a hierarchical key derivation algorithm is used to generate corresponding subkeys for each permission level. A one-way inheritance relationship is established between subkeys. For example, the subkey of system management authority can derive the subkey of operation management authority, but not vice versa. At the same time, time constraints are embedded in the subkeys. For example, the validity period of the subkey is one year. Finally, a hierarchical key derivation tree is formed, with the root node as the master key and the leaf nodes as the subkeys of each permission level, and each node has a corresponding validity period. For example, the master key generated on January 1, 2023 derives the system management key, which is valid until January 1, 2024; the system management key derives the operation management key, which is valid until January 1, 2024.

[0163] When a user terminal initiates an access request, the system performs a double identity authentication. First, the user needs to provide a username and password for identity authentication. Secondly, the system will perform a secondary verification through biometric recognition (such as fingerprint recognition or facial recognition) or behavioral feature recognition (such as keyboard input habits). After the identity verification is passed, the system and the user terminal execute the key negotiation algorithm to generate a key for this session. Use the session key to establish a two-way authentication encryption channel to ensure the confidentiality and integrity of data transmission. For example, after user B logs in with his username and password, the system requires him to perform fingerprint verification. After the verification is passed, the system negotiates with user B's terminal to generate a session key.

[0164] The system determines permissions and obtains keys based on the information of the user terminal. The system obtains the responsibility information, behavioral characteristics, and operation intentions of the user terminal and establishes an access policy model. The access policy model is mapped to the permission management contract and permission determination is performed on the blockchain. According to the determination result, the corresponding target key is obtained from the hierarchical key derivation tree. For example, user B requests access to business data, and the system verifies his permissions on the blockchain based on his operational management permissions and intention to access business data. After the verification is passed, the system obtains the operational management key from the key derivation tree.

[0165] Finally, the system decrypts and distributes the data and records the access proof. The encrypted data is decrypted using the target key to obtain the target data. The target data is distributed to the user terminal through a two-way authentication encryption channel. An access proof containing the user identity, operation record and timestamp is generated, and the access proof is written to the blockchain network to form a traceable access chain. For example, the system uses the operation management key to decrypt the market research report and sends the decrypted report to user B through an encrypted channel. At the same time, the record of user B accessing the market research report is written to the blockchain.

[0166] In this embodiment, the confidentiality and integrity of transparent management reports are effectively guaranteed through technical means such as multi-level permission system, differentiated encryption strategy, hierarchical key derivation and two-way authentication encryption channel, preventing data leakage and tampering. Through the preset permission system and access control matrix, the user's access rights can be quickly determined, and the target key can be efficiently obtained through the hierarchical key derivation tree, thereby improving data access efficiency. Writing the permission management contract and access proof into the blockchain realizes the full recording and traceability of access behavior, enhancing the audit capability of the system.

[0167] Figure 2 FIG. 1 is a schematic diagram of a transparent management system for assessment and compensation information of a new energy power plant according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0168] The first unit is used to obtain real-time power generation data, meteorological environment monitoring data and power grid dispatch instruction data of new energy power plants, use a deep learning model to analyze the meteorological environment monitoring data, predict the power generation capacity benchmark value of a specified period, build a power generation capacity assessment model based on the power generation capacity benchmark value, and calculate the comprehensive assessment benchmark value of the new energy power plant through power generation efficiency indicators and power quality indicators according to real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules;

[0169] The second unit is used to perform a hierarchical assessment of the new energy power plant based on the comprehensive assessment benchmark value, generate a power generation capacity assessment report, and use a causal analysis model to determine the cause of the power generation loss of the new energy power plant based on the power generation capacity assessment report and the power grid dispatch instruction data. When it is determined that the power generation loss is caused by the power grid dispatch restriction, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation, and the real-time power generation data, meteorological environment monitoring data, power generation capacity assessment report and compensation power data are written into the alliance blockchain network;

[0170] The third unit is used to deploy a smart contract with hierarchical permissions in the alliance blockchain network, the smart contract reads the compensation electricity data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network, generates the operation efficiency analysis result of the new energy power plant, generates a transparent management report based on the operation efficiency analysis result, the power generation capacity evaluation report, the compensation electricity data and the compensation amount, and distributes the transparent management report to user terminals with different authority levels through an encrypted channel according to the preset data access rights.

[0171] According to a third aspect of the embodiments of the present invention,

[0172] An electronic device is provided, comprising:

[0173] processor;

[0174] a memory for storing processor-executable instructions;

[0175] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0176] A fourth aspect of the embodiments of the present invention is:

[0177] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0178] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for transparent management of assessment and compensation information of new energy power plants, characterized in that: include: Acquire the real-time power generation data, meteorological environment monitoring data and power grid dispatch instruction data of the new energy power plant, use the deep learning model to analyze the meteorological environment monitoring data, predict the power generation capacity benchmark value of the specified time period, build a power generation capacity assessment model based on the power generation capacity benchmark value, and calculate the comprehensive assessment benchmark value of the new energy power plant through power generation efficiency indicators and power quality indicators according to the real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules; Based on the comprehensive assessment benchmark value, a hierarchical assessment is conducted on the new energy power plant to generate a power generation capacity assessment report. According to the power generation capacity assessment report and the grid dispatch instruction data, a causal analysis model is used to determine the cause of the power generation loss of the new energy power plant. When it is determined that the power generation loss is caused by grid dispatch restrictions, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation. The real-time power generation data, meteorological environment monitoring data, power generation capacity assessment report and compensation power data are written into the alliance blockchain network; A smart contract with hierarchical permissions is deployed in the alliance blockchain network. The smart contract reads the compensation electricity data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network, and generates an operation efficiency analysis result of the new energy power plant. A transparent management report is generated based on the operation efficiency analysis result, the power generation capacity assessment report, the compensation electricity data and the compensation amount, and the transparent management report is distributed to user terminals with different authority levels through an encrypted channel according to the preset data access rights.

2. The method according to claim 1, characterized in that Based on the power generation capacity benchmark value, a power generation capacity assessment model is constructed. According to real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules, the comprehensive assessment benchmark value of the new energy power plant is calculated through power generation efficiency indicators and power quality indicators, including: Based on the power generation capacity benchmark value, a power generation capacity assessment model including a double-layer time window is constructed, wherein the double-layer time window includes an inner time window and an outer time window, wherein the inner time window weights the short-term power generation power by a nonlinear time decay function to obtain a short-term power generation characteristic, and the outer time window weights the long-term power generation power by an exponential decreasing function to obtain a long-term power generation characteristic, and a power generation efficiency index is calculated based on a weighted combination of the short-term power generation characteristic and the long-term power generation characteristic; Acquire real-time power generation data, process the real-time power generation data using a multi-scale analysis method, obtain voltage deviation rate, three-phase voltage imbalance, frequency deviation, frequency change rate, total harmonic distortion rate and harmonic content, and normalize the processing results to construct a power quality feature vector; Inputting the power quality feature vector into a residual neural network for feature extraction, and performing attention weighting on the feature extraction result to obtain a power quality evaluation index; A hybrid evolutionary optimization algorithm is constructed, the weights of the power generation efficiency index and the power quality assessment index are encoded into real number vectors, a scaling factor is calculated based on population diversity and fitness change rate, the real number vector is adaptively mutated using the scaling factor, a differential evolution operation is combined with a particle swarm velocity update to perform a local search, the individual crowding degree is calculated by non-dominated sorting to construct a Pareto frontier, and an optimal weight vector is selected from the Pareto frontier; A three-layer fuzzy rule base is established based on preset assessment rules, the power generation efficiency index and the power quality index are evaluated in layers, the evaluation results are weighted using the optimal weight vector, and confidence correction is performed in combination with Gaussian process regression prediction to obtain a comprehensive assessment benchmark value for the new energy power plant.

3. The method according to claim 2, characterized in that A three-layer fuzzy rule base is established based on the preset assessment rules, the power generation efficiency index and the power quality index are evaluated in layers, the evaluation results are weighted by the optimal weight vector, and the confidence correction is performed in combination with Gaussian process regression prediction, and the comprehensive assessment benchmark values ​​of the new energy power plant are obtained, including: A three-layer fuzzy rule base is constructed based on the preset assessment rules, wherein the first-layer rule base contains input language variables and output language variables of the power generation efficiency index, the input language variables include five levels: ultra-low, low, medium, high and extremely high, and a Gaussian membership function is constructed based on the value range of the power generation efficiency index, and the output language variables correspond to scoring levels with different scores; The input language variables of the second-layer rule base include five levels: unqualified, basically qualified, qualified, good and excellent. A triangular membership function is constructed based on the value interval of the power quality index. The vertices and support set of the triangular membership function are dynamically optimized through a genetic algorithm. The output language variables correspond to five scoring levels and are defuzzified using the maximum membership method. The power generation efficiency index is input into the first layer rule base for fuzzy reasoning to obtain the first layer evaluation value, and the power quality index is input into the second layer rule base for fuzzy reasoning to obtain the second layer evaluation value; The third-layer rule base receives the first-layer evaluation value and the second-layer evaluation value as input, and uses a predetermined optimal weight vector to perform a weighted combination of the first-layer evaluation value and the second-layer evaluation value to obtain an initial comprehensive evaluation value; The power generation efficiency index and the power quality index form an input feature vector, the initial comprehensive evaluation value is used as an output label, a deep regression model including multiple hidden layers is constructed, an attention calculation unit is set in each hidden layer of the deep regression model to extract the correlation feature of the input feature vector, and the deep regression model is trained based on the correlation feature and the initial comprehensive evaluation value; The new power generation efficiency index and power quality index are input into the trained deep regression model to obtain the regression prediction value, the prediction error between the regression prediction value and the actual value is calculated, the confidence weight is calculated based on the prediction error, the initial comprehensive evaluation value and the regression prediction value are weighted and fused through the confidence weight to obtain the comprehensive assessment benchmark value of the new energy power plant.

4. The method according to claim 1, characterized in that: According to the power generation capacity assessment report and the grid dispatch instruction data, a causal analysis model is used to determine the cause of the power generation loss of the new energy power plant. When it is determined that the power generation loss is caused by grid dispatch restrictions, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation, including: Obtaining system parameters and grid dispatch instruction data in the power generation capacity assessment report, aligning them according to timestamps, building a causal analysis model, and using the causal analysis model to extract features from the system parameters and grid dispatch instruction data to obtain a correlation feature sequence; A dynamic Bayesian network is constructed based on the associated feature sequence, the power generation equipment state, the power grid operation state and the dispatch instruction information are used as network nodes, a state transition matrix is ​​established to characterize the dependency relationship between the network nodes, and an iterative reweighted parameter learning algorithm is used to optimize the state transition matrix to obtain the node dependency strength; The temporal causal strength is calculated based on the node dependency strength, and the dependency strength is accumulated in a sliding time window to obtain the cumulative causal strength. The temporal causal strength and the cumulative causal strength are merged by using adaptive weights to obtain the comprehensive causal strength. A piecewise decision function is constructed according to the comprehensive causal strength to determine whether the power generation loss is caused by grid dispatch restrictions; When it is determined that the power generation loss is caused by the grid dispatch restriction, a power generation equipment state vector is constructed based on the associated characteristic sequence, the power generation equipment state vector and the dependency strength are input into a state predictor based on a high-order Taylor series, a power generation capacity benchmark model is established according to an output result of the state predictor, a correction coefficient is calculated using a recursive residual compensation algorithm based on the dependency strength, and the correction coefficient is multiplied by the output value of the power generation capacity benchmark model to obtain a theoretical power generation; A compensation optimization objective function is constructed according to the deviation between the theoretical power generation and the actual power generation, a dynamic relaxation factor is introduced into the compensation optimization objective function to adjust the compensation intensity, an improved gradient projection algorithm is used to solve the compensation coefficient, and the product of the compensation coefficient and the lost electricity is determined as the final compensation electricity data of the new energy power plant.

5. The method according to claim 4, characterized in that Inputting the state vector of the power generation equipment and the dependency strength into a state predictor based on a high-order Taylor series, establishing a power generation capacity benchmark model according to the output result of the state predictor, calculating a correction coefficient using a recursive residual compensation algorithm based on the dependency strength, and multiplying the correction coefficient by the output value of the power generation capacity benchmark model to obtain a theoretical power generation includes: Decompose the state vector and the dependency strength of the power generation equipment by wavelet transform to obtain the first characteristic component and the second characteristic component, determine the Taylor series expansion order corresponding to each characteristic component based on the minimum description length criterion, and perform Taylor series expansion on the first characteristic component and the state mapping function of the first characteristic component to obtain the first characteristic prediction sequence and the second characteristic prediction sequence; Projecting the dependency intensity matrix onto a Riemann manifold space to obtain a geometric eigenvector, constructing a dynamic coupling weight matrix based on the geometric eigenvector, and using the dynamic coupling weight matrix to perform a weighted combination of the first feature prediction sequence and the second feature prediction sequence to obtain a state prediction vector; Input the state prediction vector into a pre-built deep probability model to extract probability distribution features, perform multi-head time series attention calculation on the probability distribution features to obtain time series correlation features, build a power generation capacity benchmark model based on the time series correlation features, and calculate the benchmark power generation through the power generation capacity benchmark model; Projecting the difference sequence between the reference power generation and the actual power generation into the orthogonal wavelet basis space to obtain multi-level residual components, constructing residual compensation coefficients based on the characteristic decomposition results of the dependency intensity, and using the residual compensation coefficients to dynamically compensate the multi-level residual components; Constructing a graph attention compensation network, inputting the compensated multi-level residual components into the graph attention compensation network, obtaining a dynamic correction factor sequence through recursive operation, and establishing a time series consistency evaluation model at the same time, multiplying the dynamic correction factor sequence with the benchmark power generation to obtain a candidate theoretical power generation, and calculating the distribution deviation between the candidate theoretical power generation and the actual power generation based on the time series consistency evaluation model; An adaptive optimizer based on Riemann metric is constructed, and the adaptive optimizer is used to jointly optimize the state prediction vector, the benchmark power generation and the dynamic correction factor sequence, so as to minimize the distribution deviation and obtain the final theoretical power generation.

6. The method according to claim 1, characterized in that A smart contract with hierarchical permissions is deployed in the alliance blockchain network. The smart contract reads the compensation power data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power generation conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network. The operation efficiency analysis results of the new energy power plant are generated, including: Deploy a hierarchical permission smart contract with management layer, execution layer and query layer in the alliance blockchain network, where the management layer uses asymmetric keys for identity authentication, the execution layer sets the compensation rule state machine, the query layer configures data access rights, and generates a contract deployment identifier; Input the contract deployment identifier and the compensation power data into the compensation rule state machine, calculate the compensation benchmark value based on the segmented mapping function, use a deep neural network to predict the compensation adjustment factor, and multiply the compensation benchmark value by the compensation adjustment factor to obtain an initial compensation amount; The initial compensation amount and the historical compensation record form a compensation state vector, a temporal attention network is used to extract the temporal features of the compensation state vector, a compensation constraint model is constructed based on the temporal features, and the compensation amount that satisfies the compensation constraint model is written into the alliance blockchain network; Read the operation data of the new energy power plant from the alliance blockchain network and perform wavelet decomposition to obtain an operation characteristic sequence, calculate the power generation conversion efficiency index based on the operation characteristic sequence, and calculate the equipment utilization index based on the equipment operation time to obtain an efficiency index set; Input the efficiency index set into the graph convolutional neural network to extract spatial correlation features, build a new energy power plant efficiency evaluation model and calculate the operating efficiency numerical sequence; The operating efficiency numerical sequence is segmented into sliding time windows, the efficiency change trend in each time window is calculated using the exponential smoothing method, a power plant similarity matrix is ​​constructed based on the efficiency change trend, and the power plants are grouped using a density clustering algorithm; The probability distribution of the operating efficiency of each power plant group is calculated, the degree of deviation is identified using the local anomaly factor algorithm, efficiency anomaly features are generated, the efficiency change trend and the efficiency anomaly features are input into the deep neural network to generate the operating efficiency analysis results of the new energy power plant.

7. The method according to claim 1, characterized in that Distributing the transparent management report to user terminals of different authority levels through an encrypted channel according to preset data access rights includes: Construct a multi-level permission system including system management permission, operation management permission, technical personnel permission, partner permission and public permission, set up an access control matrix based on the multi-level permission system, generate a permission management contract using identity authentication information and timestamp, and deploy the permission management contract to the blockchain network; The transparent management report is divided into core data, commercial data and public data according to data sensitivity, and encrypted using differentiated encryption strategies such as group encryption, block encryption and stream encryption to obtain encrypted data; Generate a master key, generate a hierarchical subkey sequence for the multi-level authority system based on the master key using a hierarchical key derivation algorithm, establish a unidirectional inheritance relationship between the hierarchical subkey sequences and embed a time constraint condition, and construct a hierarchical key derivation tree; Receive an access request sent by a user terminal, verify the identity of the user terminal using dual identity authentication and behavior feature recognition, perform key negotiation with the authenticated user terminal to generate a session key, and establish a two-way authentication encryption channel based on the session key; Obtaining the responsibility information, behavior characteristics and operation intention of the user terminal, establishing an access policy model, mapping the access policy model to a rights management contract for rights determination, and obtaining a target key from the hierarchical key derivation tree based on the determination result; The encrypted data is decrypted using the target key to obtain the target data, the target data is distributed to the user terminal through the two-way authentication encryption channel, and an access certificate including an identity, an operation record and a timestamp is generated, and the access certificate is written into the blockchain network to form a traceable access chain.

8. A transparent management system for assessment and compensation information of new energy power plants, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to obtain real-time power generation data, meteorological environment monitoring data and power grid dispatch instruction data of new energy power plants, use a deep learning model to analyze the meteorological environment monitoring data, predict the power generation capacity benchmark value of a specified period, build a power generation capacity assessment model based on the power generation capacity benchmark value, and calculate the comprehensive assessment benchmark value of the new energy power plant through power generation efficiency indicators and power quality indicators according to real-time power generation data, combined with the power generation capacity assessment model and preset assessment rules; The second unit is used to perform a hierarchical assessment of the new energy power plant based on the comprehensive assessment benchmark value, generate a power generation capacity assessment report, and use a causal analysis model to determine the cause of the power generation loss of the new energy power plant based on the power generation capacity assessment report and the power grid dispatch instruction data. When it is determined that the power generation loss is caused by the power grid dispatch restriction, the theoretical power generation is calculated based on the power generation capacity assessment model, and the compensation power data of the new energy power plant is calculated in combination with the actual power generation, and the real-time power generation data, meteorological environment monitoring data, power generation capacity assessment report and compensation power data are written into the alliance blockchain network; The third unit is used to deploy a smart contract with hierarchical permissions in the alliance blockchain network, the smart contract reads the compensation electricity data, automatically calculates the compensation amount according to the preset compensation rules, and calculates the power conversion efficiency index and equipment utilization index of the new energy power plant based on the historical data in the alliance blockchain network, generates the operation efficiency analysis result of the new energy power plant, generates a transparent management report based on the operation efficiency analysis result, the power generation capacity evaluation report, the compensation electricity data and the compensation amount, and distributes the transparent management report to user terminals with different authority levels through an encrypted channel according to the preset data access rights.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images