Cooperative operation and maintenance method and system of energy storage power station

Abstract

The application discloses a kind of collaborative operation methods of energy storage power station, including obtaining the data information of target energy storage power station and preprocessing to construct the panoramic real-time dataset of target energy storage power station;According to the panoramic real-time dataset obtained, the digital twin model is trained, and the distributed collaborative digital twin model of target energy storage power station is obtained;According to the distributed collaborative digital twin model obtained, the distributed collaborative analysis is carried out to target energy storage power station, and the global operation state evaluation of target energy storage power station is realized;According to the evaluation result obtained, the index of target energy storage power station is predicted based on neural network model, and the prediction result is corrected;According to the data information obtained, operation and maintenance decision result is generated, and the collaborative operation of target energy storage power station is completed.The application also discloses a kind of systems for realizing the collaborative operation method of the energy storage power station.The application not only realizes the collaborative operation of energy storage power station, but also has higher reliability and better accuracy.

Description

Technical Field

[0001] This invention belongs to the field of electrical automation, specifically relating to a collaborative operation and maintenance method and system for energy storage power stations. Background Technology

[0002] With economic and technological development and the improvement of people's living standards, electricity has become an indispensable secondary energy source in people's production and daily life, bringing endless convenience. Therefore, ensuring a stable and reliable supply of electricity has become one of the most important tasks of the power system.

[0003] Currently, an increasing number of new energy power generation systems are being integrated into the power grid and generating electricity. The randomness and intermittency of the output of these new energy power generation systems pose a significant challenge to the safe and stable operation of the power system. To address this challenge, energy storage systems are beginning to be widely used in power systems.

[0004] However, current energy storage system operation and maintenance solutions often employ single-mode solutions based on static models. These solutions lack coordination capabilities and struggle to meet the demands for high-precision, low-latency, and distributed operation and maintenance under complex operating conditions. Summary of the Invention

[0005] One of the objectives of this invention is to provide a collaborative operation and maintenance method for energy storage power stations that is highly reliable and accurate.

[0006] The second objective of this invention is to provide a system for implementing the collaborative operation and maintenance method of the energy storage power station.

[0007] The collaborative operation and maintenance method for energy storage power stations provided by this invention includes the following steps:

[0008] S1. Obtain data information from the target energy storage power station;

[0009] S2. Preprocess the data information obtained in step S1 to construct a panoramic real-time dataset of the target energy storage power station;

[0010] S3. Based on the panoramic real-time dataset obtained in step S2, train the digital twin model to obtain the distributed collaborative digital twin model of the target energy storage power station;

[0011] S4. Based on the distributed collaborative digital twin model obtained in step S3, perform distributed collaborative analysis on the target energy storage power station and achieve a global operational status assessment of the target energy storage power station;

[0012] S5. Based on the evaluation results obtained in step S4, predict the indicators of the target energy storage power station using a neural network model, and correct the prediction results.

[0013] S6. Based on the data obtained in step S5, generate operation and maintenance decision results to complete the collaborative operation and maintenance of the target energy storage power station.

[0014] Step S1, which involves acquiring data information of the target energy storage power station, specifically includes the following steps:

[0015] Obtain data information from the target energy storage power station;

[0016] The data information includes data information of the battery cluster, data information of the battery management system, data information of the energy storage converter, and environmental data information of the target energy storage power station;

[0017] The data information of the battery cluster includes voltage data information, current data information and temperature data information;

[0018] The data information of the battery management system includes cell voltage data, SOC data, and SOH data;

[0019] The data information of the energy storage converter includes power data, DC side voltage data, and efficiency data.

[0020] Step S2 involves preprocessing the data information obtained in step S1 to construct a panoramic real-time dataset of the target energy storage power station. This process specifically includes the following steps:

[0021] The data obtained in step S1 is processed using the following formula:

[0022] In the formula The processed measurement value acquired by the i-th sensor at time t; This represents the raw measurement value acquired by the i-th sensor at time tk; The weighting factor is set. This represents the number of data points within the observation window. This represents the average value of the measurements taken by the i-th sensor within the set observation window. Let be the standard deviation of the measurements taken by the i-th sensor within the set observation window;

[0023] For device j, construct the corresponding original protocol features. for ,in For equipment type identification, For data frame structure description; set of state parameters Set as , Let L be the measured value of the Lth state variable output by device j;

[0024] according to and The original bitstream is unpacked into a state vector with physical units. for ;

[0025] Using the following formula, Convert each component in the equation to a dimensionless standardized value. :

[0026] In the formula This represents the mean of the corresponding components; This represents the standard deviation of the corresponding component;

[0027] All Constructing a standardized data descriptor for ; It is used to enable data exchange between various brands and models of sensors.

[0028] Step S3, which involves training the digital twin model based on the panoramic real-time dataset obtained in step S2 to obtain a distributed collaborative digital twin model of the target energy storage power station, specifically includes the following steps:

[0029] Sub-models are established for the battery clusters, battery management system and energy storage converter in the target energy storage system, respectively;

[0030] Constructing a battery cluster sub-model:

[0031] The battery cluster model adopts a first-order RC equivalent circuit model; the state equation of the model is expressed as:

[0032] In the formula Let be the terminal voltage of the battery cluster at time t; Let be the current of the battery cluster at time t; The internal resistance of the battery cluster is ohmic. The polarization internal resistance of the battery cluster; The data sampling interval; It is a time constant; This is the open-circuit voltage of the battery cluster;

[0033] Constructing a sub-model for the battery management system:

[0034] A sub-model of the battery management system is constructed using a gated recurrent unit neural network;

[0035] The forward propagation formula for the model is:

[0036] In the formula Let be the hidden state at time t; For gated loop unit; Input data for the battery management system sub-model; The SOC state value predicted by the model at time t; The weight vector for SOC prediction; The bias for SOC prediction; The SOH state value predicted by the model at time t; The weight vector predicted by SOH; The bias for SOH prediction; This is the softplus function;

[0037] Construct a sub-model for the energy storage converter:

[0038] The energy storage converter sub-model is constructed using the following formula:

[0039] In the formula AC side power; Let be the efficiency of the energy storage converter corresponding to the kth discrete operating point; This represents the DC-side power corresponding to the kth discrete operating point. DC-side power;

[0040] For each constructed sub-model, the parameters of the sub-model are updated using the following formula:

[0041] In the formula The parameter values ​​of sub-model i at time t+1; Let be the parameter values ​​of sub-model i at time t; The set learning rate; Let be the predicted output of sub-model i at time t; This represents the actual output of sub-model i at time t; This is the residual loss function; The gradient of the residual loss function;

[0042] During training of each sub-model, the following globally consistent loss is constructed:

[0043] In the formula The global consistency loss value is denoted by n; n is the number of sub-models. The weights of the i-th sub-model are set; For global reference state, , This is the core prediction output of the i-th sub-model at time t; The calculation process is used to eliminate the dimensional differences in the outputs of different sub-models and form a global benchmark that fits the overall operating characteristics of the power plant.

[0044] By using global consistency loss, the sub-model can maintain consistency with the overall operating characteristics while fitting its own data.

[0045] Step S4, which involves performing distributed collaborative analysis on the target energy storage power station based on the distributed collaborative digital twin model obtained in step S3, and achieving a global operational status assessment of the target energy storage power station, specifically includes the following steps:

[0046] Constructing a global runtime state vector :

[0047] In the formula The cumulative charge and discharge capacity of the target energy storage power station. , The total number of battery clusters within the target energy storage power station. This is the current time step for statistics. For battery cluster b at time DC power, The sampling interval; This represents the maximum charge / discharge power in the hourly range. , Indicates the most recent hour. The charging and discharging power of the target energy storage power station; For the current power, the continuous operating time , Let b be the state of charge of the b-th battery cluster at time t. The rated capacity of the b-th battery cluster is... For a given minimum number; The weighted average SOC value. ; Minimum SOH value; global running state vector It is used to fully characterize the global operating status of energy storage power stations and is the basic definition of the system's operating status. It is used to support the integrity of the scheme and subsequent multi-dimensional status assessment and expansion.

[0048] The following formula is used as the comprehensive evaluation function:

[0049] In the formula To comprehensively evaluate the function value; The set credibility weight; The SOC contribution coefficient is set. The contribution coefficient of SOH is set;

[0050] Based on the obtained comprehensive evaluation function value, the consistency deviation is calculated. for ;

[0051] Consistency deviation Make a judgment:

[0052] like If the value exceeds a set threshold, then sub-model i is retrained; simultaneously, adjustments are made. The value is , The set attenuation coefficient.

[0053] Step S5, which involves predicting the indicators of the target energy storage power station based on the evaluation results obtained in step S4 using a neural network model and then correcting the prediction results, specifically includes the following steps:

[0054] Constructing key indicator vectors for ;in For maximum charge and discharge power, Let be the energy conversion efficiency at time t;

[0055] The input to the neural network model is The output is a future time. Key Indicator Vector Predicted value ;

[0056] The neural network model employs a three-layer long short-term memory network structure; the LSTM layer updates the memory units. and hidden state The formula for calculating network forward propagation is:

[0057] In the formula The output of the input gate; It is the sigmoid activation function; Here is the input weight matrix of the input gate; Let be the hidden state weight matrix of the input gate; For the input gate bias; The output of the forget gate; This is the input weight matrix for the forget gate; Let be the hidden state weight matrix of the forget gate; For the offset of the forget gate; This is the output of the output gate; This is the input weight matrix for the output gate; Let be the hidden state weight matrix of the output gate; For the output gate bias; The candidate state of the cell at time t; The input weight matrix represents the candidate cell states; The hidden state weight matrix represents the candidate cell state. This is a bias term for the candidate state of the cell; The cell state at time t; This is element-wise multiplication;

[0058] Input the hidden state of the last time step into the fully connected output layer to obtain the future time step. Key Indicator Vector Predicted value for ;in To output the weight matrix, For output bias;

[0059] The residual value was calculated. for ;

[0060] When the residual value When the value is less than or equal to the set threshold, the output of the sub-model is directly used as the final result;

[0061] When the residual value When the value exceeds the set threshold, the final result is calculated using the following formula:

[0062] In the formula This is the final result obtained from the calculation; The set weight value.

[0063] Step S6, which involves generating operation and maintenance decision results based on the data information obtained in step S5 to complete the collaborative operation and maintenance of the target energy storage power station, specifically includes the following steps:

[0064] Based on the data obtained in step S5, an optimization model is constructed to adjust the operation mode of the target energy storage power station;

[0065] The following formula is used as the objective function of the optimization model:

[0066] In the formula The objective function value; This is the first weighting coefficient set; The actual charging and discharging power at time k; This is a reference value for charging and discharging power. This is the second weighting coefficient that is set; The energy conversion efficiency of the energy storage power station; This is the set third weighting coefficient; As a risk factor, , The weighting coefficient for health status. The SOH value of the k1th battery cluster is... Temperature weighting coefficient, This represents the highest cell temperature in the battery cluster. The set safe temperature threshold;

[0067] The following formula is used as the constraint condition for the optimization model:

[0068] In the formula The charging power of the target energy storage power station at time t; This is the maximum charging power; Let be the discharge power of the target energy storage power station at time t; This represents the maximum discharge power. Minimum SOC value; The maximum SOC value; This is the minimum allowable operating temperature for the cooling system; The actual operating temperature of the cooling system at time t; This is the maximum allowable operating temperature of the cooling system;

[0069] The constructed optimization model is solved to obtain decision variables, which are then used as the operation and maintenance decision results; the decision variables include , , and the fault status indicator variable of the energy storage power station at time t ;

[0070] The operation and maintenance decisions are used to complete the collaborative operation and maintenance of the target energy storage power station.

[0071] Step S6 further includes the following steps:

[0072] The instruction execution consistency value is calculated using the following formula:

[0073] In the formula The instruction execution consistency value for subsystem i; The target value of the instruction for subsystem i; This is the rated reference value of the instruction control quantity corresponding to subsystem i;

[0074] Determine the consistency of instruction execution values:

[0075] If the instruction execution consistency value is greater than or equal to the set threshold, the subsystem execution accuracy is determined to meet the requirements.

[0076] If the instruction execution consistency value is less than the set threshold, it is determined that the subsystem's execution accuracy does not meet the requirements. At this time, the instruction target value is sent to the corresponding subsystem again, and monitoring is performed again.

[0077] This invention also provides a system for implementing the collaborative operation and maintenance method of the energy storage power station, comprising a data acquisition module, a data processing module, a model building module, a status assessment module, a result correction module, and a collaborative operation and maintenance module; the data acquisition module, data processing module, model building module, status assessment module, result correction module, and collaborative operation and maintenance module are connected in series; the data acquisition module is used to acquire data information of the target energy storage power station and upload the data information to the data processing module; the data processing module is used to preprocess the acquired data information according to the received data information to construct a panoramic real-time dataset of the target energy storage power station and upload the data information to the model building module; the model building module is used to, according to the received data information and the obtained panoramic real-time dataset, modify the digital twin model... The system trains a distributed collaborative digital twin model of the target energy storage power station to obtain the data information, which is then uploaded to the status assessment module. The status assessment module performs distributed collaborative analysis of the target energy storage power station based on the received data and the obtained distributed collaborative digital twin model, achieving a global operational status assessment of the target energy storage power station, and uploads the data information to the result correction module. The result correction module predicts the indicators of the target energy storage power station based on the received data and the assessment results using a neural network model, corrects the prediction results, and uploads the data information to the collaborative operation and maintenance module. The collaborative operation and maintenance module generates operation and maintenance decision results based on the received data and the obtained data, completing the collaborative operation and maintenance of the target energy storage power station.

[0078] The collaborative operation and maintenance method and system for energy storage power stations provided by this invention acquires and processes various data information of the target energy storage power station, and makes global decisions and corrections based on a digital twin model. This not only achieves collaborative operation and maintenance of energy storage power stations, but also has higher reliability and better accuracy. Attached Figure Description

[0079] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0080] Figure 2 This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation

[0081] like Figure 1 The diagram shown is a flowchart of the method of the present invention: The collaborative operation and maintenance method for energy storage power stations disclosed in this invention includes the following steps:

[0082] S1. Obtain data information from the target energy storage power station; specifically including the following steps:

[0083] Obtain data information from the target energy storage power station;

[0084] The data information includes data information of the battery cluster, data information of the battery management system, data information of the energy storage converter, and environmental data information of the target energy storage power station;

[0085] The data information of the battery cluster includes voltage data information, current data information and temperature data information;

[0086] The data information of the battery management system includes cell voltage data, SOC data, and SOH data;

[0087] The data information of the energy storage converter includes power data, DC side voltage data, and efficiency data.

[0088] In practice, various types of sensors can be used to acquire the above data information;

[0089] S2. Preprocess the data information obtained in step S1 to construct a panoramic real-time dataset of the target energy storage power station; specifically including the following steps:

[0090] The data obtained in step S1 is processed using the following formula:

[0091] In the formula The processed measurement value acquired by the i-th sensor at time t; This represents the raw measurement value acquired by the i-th sensor at time tk; The weighting factor is set. This represents the number of data points within the observation window. This represents the average value of the measurements taken by the i-th sensor within the set observation window. denoted as the standard deviation of the measurements taken by the i-th sensor within the set observation window; the processed results can significantly reduce the influence of noise and outliers while preserving the true dynamic change trend, and maintain the consistency of data distribution among different sensors;

[0092] In practice, due to differences in data protocols among sensor manufacturers, the following steps are taken to unify them:

[0093] For device j, construct the corresponding original protocol features. for ,in For equipment type identification, For data frame structure description; set of state parameters Set as , Let L be the measured value of the Lth state variable output by device j;

[0094] according to and The original bitstream is unpacked into a state vector with physical units. for ;

[0095] Using the following formula, Convert each component in the equation to a dimensionless standardized value. :

[0096] In the formula This represents the mean of the corresponding components; This represents the standard deviation of the corresponding component;

[0097] All Constructing a standardized data descriptor for ; Used to enable data exchange between various brands and models of sensors;

[0098] S3. Based on the panoramic real-time dataset obtained in step S2, train the digital twin model to obtain a distributed collaborative digital twin model of the target energy storage power station; specifically including the following steps:

[0099] Sub-models are established for the battery clusters, battery management system and energy storage converter in the target energy storage system, respectively;

[0100] Constructing a battery cluster sub-model:

[0101] The battery cluster model adopts a first-order RC equivalent circuit model; the state equation of the model is expressed as:

[0102] In the formula Let be the terminal voltage of the battery cluster at time t; Let be the current of the battery cluster at time t; The internal resistance of the battery cluster is ohmic. The polarization internal resistance of the battery cluster; The data sampling interval; It is a time constant; This is the open-circuit voltage of the battery cluster;

[0103] Constructing a sub-model for the battery management system:

[0104] A sub-model of the battery management system is constructed using a gated recurrent unit neural network;

[0105] The forward propagation formula for the model is:

[0106] In the formula Let be the hidden state at time t; For gated loop unit; Input data for the battery management system sub-model; The SOC state value predicted by the model at time t; The weight vector for SOC prediction; The bias for SOC prediction; The SOH state value predicted by the model at time t; The weight vector predicted by SOH; The bias for SOH prediction; This is the softplus function;

[0107] Construct a sub-model for the energy storage converter:

[0108] The energy storage converter sub-model is constructed using the following formula:

[0109] In the formula AC side power; Let be the efficiency of the energy storage converter corresponding to the kth discrete operating point; This represents the DC-side power corresponding to the kth discrete operating point. DC-side power;

[0110] For each constructed sub-model, the parameters of the sub-model are updated using the following formula:

[0111] In the formula The parameter values ​​of sub-model i at time t+1; Let be the parameter values ​​of sub-model i at time t; The set learning rate; Let be the predicted output of sub-model i at time t; This represents the actual output of sub-model i at time t; This is the residual loss function; The gradient of the residual loss function is given. The above model ensures that each sub-model remains consistent with the real-time operating data, thereby improving the model's ability to represent the nonlinear characteristics and sudden operating conditions of the system.

[0112] During training of each sub-model, the following globally consistent loss is constructed:

[0113] In the formula The global consistency loss value is denoted by n; n is the number of sub-models. The weights of the i-th sub-model are set; For global reference state, , This is the core prediction output of the i-th sub-model at time t; The calculation process is used to eliminate the dimensional differences in the outputs of different sub-models and form a global benchmark that fits the overall operating characteristics of the power plant.

[0114] By using global consistency loss, the sub-model can maintain consistency with the overall operating characteristics while fitting its own data, ultimately forming a digital twin model that can be dynamically reconstructed in a distributed environment.

[0115] S4. Based on the distributed collaborative digital twin model obtained in step S3, perform distributed collaborative analysis on the target energy storage power station and achieve a global operational status assessment of the target energy storage power station; specifically, this includes the following steps:

[0116] Constructing a global runtime state vector :

[0117] In the formula The cumulative charge and discharge capacity of the target energy storage power station. , The total number of battery clusters within the target energy storage power station. This is the current time step for statistics. For battery cluster b at time DC power, The sampling interval; This represents the maximum charge / discharge power in the hourly range. , Indicates the most recent hour. The charging and discharging power of the target energy storage power station; For the current power, the continuous operating time , Let b be the state of charge of the b-th battery cluster at time t. The rated capacity of the b-th battery cluster is... For a given minimum number; The weighted average SOC value. ; Minimum SOH value; global running state vector It is used to fully characterize the global operating status of energy storage power stations and is the basic definition of the system's operating status. It is used to support the integrity of the scheme and subsequent multi-dimensional status assessment and expansion.

[0118] The following formula is used as the comprehensive evaluation function:

[0119] In the formula To comprehensively evaluate the function value; The set credibility weight; The SOC contribution coefficient is set. The contribution coefficient of SOH is set; this comprehensive evaluation function ensures that the operating characteristics of individual equipment are reflected in the fusion process, while preserving the overall system health, thereby realizing the characterization of the operating status at the power plant level;

[0120] Based on the obtained comprehensive evaluation function value, the consistency deviation is calculated. for ;

[0121] Consistency deviation Make a judgment:

[0122] like If the value exceeds a set threshold, then sub-model i is retrained; simultaneously, adjustments are made. The value is , The set attenuation coefficient;

[0123] This mechanism can weaken the influence of mismatch sub-models in global decision-making and promote their rapid correction, thereby improving the robustness and effectiveness of the overall evaluation results.

[0124] S5. Based on the evaluation results obtained in step S4, predict the indicators of the target energy storage power station using a neural network model, and correct the prediction results; specifically, this includes the following steps:

[0125] Constructing key indicator vectors for ;in For maximum charge and discharge power, Let be the energy conversion efficiency at time t;

[0126] The input to the neural network model is The output is a future time. Key Indicator Vector Predicted value ;

[0127] The neural network model employs a three-layer long short-term memory network structure; the LSTM layer updates the memory units. and hidden state The formula for calculating network forward propagation is:

[0128] In the formula The output of the input gate; It is the sigmoid activation function; Here is the input weight matrix of the input gate; Let be the hidden state weight matrix of the input gate; For the input gate bias; The output of the forget gate; This is the input weight matrix for the forget gate; Let be the hidden state weight matrix of the forget gate; For the offset of the forget gate; This is the output of the output gate; This is the input weight matrix for the output gate; Let be the hidden state weight matrix of the output gate; For the output gate bias; The candidate state of the cell at time t; The input weight matrix represents the candidate cell states; The hidden state weight matrix represents the candidate cell state. This is a bias term for the candidate state of the cell; The cell state at time t; This is element-wise multiplication;

[0129] Input the hidden state of the last time step into the fully connected output layer to obtain the future time step. Key Indicator Vector Predicted value for ;in To output the weight matrix, For output bias;

[0130] The residual value was calculated. for ;

[0131] When the residual value When the value is less than or equal to the set threshold, the output of the sub-model is directly used as the final result;

[0132] When the residual value When the value exceeds the set threshold, the final result is calculated using the following formula:

[0133] In the formula This is the final result obtained from the calculation; The set weight value;

[0134] S6. Based on the data obtained in step S5, generate operation and maintenance decision results to complete the collaborative operation and maintenance of the target energy storage power station; specifically including the following steps:

[0135] Based on the data obtained in step S5, an optimization model is constructed to adjust the operation mode of the target energy storage power station;

[0136] The following formula is used as the objective function of the optimization model:

[0137] In the formula The objective function value; This is the first weighting coefficient set; The actual charging and discharging power at time k; This is a reference value for charging and discharging power. This is the second weighting coefficient that is set; The energy conversion efficiency of the energy storage power station; This is the set third weighting coefficient; As a risk factor, , The weighting coefficient for health status. The SOH value of the k1th battery cluster is... Temperature weighting coefficient, This represents the highest cell temperature in the battery cluster. The set safe temperature threshold;

[0138] The following formula is used as the constraint condition for the optimization model:

[0139] In the formula The charging power of the target energy storage power station at time t; This is the maximum charging power; Let be the discharge power of the target energy storage power station at time t; This represents the maximum discharge power. Minimum SOC value; The maximum SOC value; This is the minimum allowable operating temperature for the cooling system; The actual operating temperature of the cooling system at time t; This is the maximum allowable operating temperature of the cooling system;

[0140] This optimization function achieves comprehensive optimization of the power plant in terms of economy, safety and reliability by balancing power tracking accuracy, energy efficiency level and potential risks, and finally outputs the optimal decision variables.

[0141] The constructed optimization model is solved to obtain decision variables, which are then used as the operation and maintenance decision results; the decision variables include , , and the fault status indicator variable of the energy storage power station at time t ;

[0142] The operation and maintenance decisions are used to complete the collaborative operation and maintenance of the target energy storage power station;

[0143] In practice, the following formula is also used to calculate the instruction execution consistency value:

[0144] In the formula The instruction execution consistency value for subsystem i; The target value of the instruction for subsystem i; This is the rated reference value of the instruction control quantity corresponding to subsystem i;

[0145] Determine the consistency of instruction execution values:

[0146] If the instruction execution consistency value is greater than or equal to the set threshold, the subsystem execution accuracy is determined to meet the requirements.

[0147] If the instruction execution consistency value is less than the set threshold, it is determined that the subsystem's execution accuracy does not meet the requirements. At this time, the instruction target value is sent to the corresponding subsystem again, and monitoring is performed again.

[0148] The method of the present invention will be further described below with reference to an embodiment:

[0149] A distributed energy storage power station with a rated power of 1MW and a rated capacity of 2MWh was selected as the test object. Actual operating data was collected for 30 consecutive days. The operation and maintenance assessment and decision-making were conducted using the method of this invention and the following two comparative methods:

[0150] The first comparison method is an energy storage operation and maintenance method based on cloud-based centralized digital twins. Its model parameters are updated offline once a week and do not include sub-model collaboration mechanism and online real-time correction.

[0151] In contrast, Method 2 is a conventional operation and maintenance method based on a static equivalent circuit model, which relies solely on periodic manual inspections of the SOC and SOH for scheduling.

[0152] The comparison metrics include: the average prediction error of the digital twin model (for terminal voltage, SOC, and SOH), the root mean square error between the condition assessment and the actual value, the average delay time of operation and maintenance decision response, and the overall operating efficiency of the power plant throughout its entire lifecycle.

[0153] Under identical hardware conditions, the test results show that: the average absolute error of the model prediction terminal voltage using the method of this invention is 0.85%, while that of comparative method one is 2.31% and comparative method two is 4.67%; the root mean square error of SOC estimation is 1.12% for this invention, 3.08% for comparative method one, and 6.54% for comparative method two; the root mean square error of SOH estimation is 0.78% for this invention, 2.15% for comparative method one, and 5.23% for comparative method two; the operation and maintenance decision response latency is 87 milliseconds for this invention, 460 milliseconds for comparative method one (due to cloud communication and batch calculation), and more than 10 seconds for comparative method two (relying on manual intervention); the overall operating efficiency is 89.6% for this invention, 83.2% for comparative method one, and 74.5% for comparative method two. These comparative data demonstrate that this invention, through an event-driven distributed collaborative mechanism, a dynamically reconstructed digital twin model, and a real-time error correction strategy, significantly improves model fidelity, real-time response, and overall system operating efficiency, overcoming the shortcomings of existing technologies such as model accuracy decay over time, insufficient collaborative capabilities, and high latency.

[0154] This invention achieves unified access and standardized processing of heterogeneous data from multiple sources, such as electrical, chemical, and thermal data, through distributed data acquisition within energy storage power stations and by combining time-series consistency correction and semantic mapping functions. This ensures the accuracy and semantic consistency of the panoramic real-time dataset, laying a solid foundation for subsequent digital twin modeling.

[0155] The present invention constructs a distributed collaborative digital twin model and introduces parameterized mapping and cross-sub-model consistency constraints to achieve dynamic reconstruction and online training of the model under complex time-varying conditions, enabling it to maintain high fidelity and robustness continuously, and avoiding the failure problem of traditional static modeling in long-term operation.

[0156] The present invention integrates and constrains the prediction results of local sub-models with global operating characteristics by designing a comprehensive state evaluation function and a consistency deviation correction mechanism. This enables the coordinated characterization of multi-level operating characteristics of battery clusters, battery management systems (BMS), and energy storage converters (PCS) and the accurate assessment of global operating status, effectively improving the reliability of state monitoring and health management.

[0157] The present invention forms a dynamic residual constraint and credibility factor-driven correction mechanism by weighted fusion of real-time error correction and neural network prediction results. This ensures the balance between calculated and predicted values ​​and realizes the long-term stability and generalization ability of digital twins in non-stationary operating environments, thereby overcoming the problem of the accuracy of traditional models decaying over time.

[0158] The present invention constructs a multi-objective optimization function and a distributed execution consistency verification mechanism to transform the global state evaluation results into precise operation and maintenance decisions, and ensures the implementation of the decisions through closed-loop feedback, thus forming an adaptive and self-optimizing distributed operation and maintenance system, which effectively improves the economy, safety and reliability of energy storage power stations.

[0159] like Figure 2The diagram shows the functional modules of the system of the present invention: The system for implementing the collaborative operation and maintenance method of the energy storage power station disclosed in this invention includes a data acquisition module, a data processing module, a model building module, a status assessment module, a result correction module, and a collaborative operation and maintenance module; the data acquisition module, data processing module, model building module, status assessment module, result correction module, and collaborative operation and maintenance module are connected in series; the data acquisition module is used to acquire data information of the target energy storage power station and upload the data information to the data processing module; the data processing module is used to preprocess the acquired data information according to the received data information to construct a panoramic real-time dataset of the target energy storage power station and upload the data information to the model building module; the model building module is used to preprocess the acquired data information according to the received data information and the obtained panoramic real-time dataset. The system trains a digital twin model to obtain a distributed collaborative digital twin model of the target energy storage power station and uploads the data to the status assessment module. The status assessment module performs distributed collaborative analysis of the target energy storage power station based on the received data and the obtained distributed collaborative digital twin model, achieving a global operational status assessment of the target energy storage power station, and uploads the data to the result correction module. The result correction module predicts the indicators of the target energy storage power station based on the received data and the assessment results using a neural network model, corrects the prediction results, and uploads the data to the collaborative operation and maintenance module. The collaborative operation and maintenance module generates operation and maintenance decision results based on the received data and the obtained data, completing the collaborative operation and maintenance of the target energy storage power station.

Claims

1. A collaborative operation and maintenance method for an energy storage power station, comprising the following steps: S1. Obtain data information from the target energy storage power station; S2. Preprocess the data information obtained in step S1 to construct a panoramic real-time dataset of the target energy storage power station; S3. Based on the panoramic real-time dataset obtained in step S2, train the digital twin model to obtain the distributed collaborative digital twin model of the target energy storage power station; S4. Based on the distributed collaborative digital twin model obtained in step S3, perform distributed collaborative analysis on the target energy storage power station and achieve a global operational status assessment of the target energy storage power station; S5. Based on the evaluation results obtained in step S4, predict the indicators of the target energy storage power station using a neural network model, and correct the prediction results. S6. Based on the data obtained in step S5, generate operation and maintenance decision results to complete the collaborative operation and maintenance of the target energy storage power station.

2. The collaborative operation and maintenance method for energy storage power stations according to claim 1, characterized in that... Step S1, which involves acquiring data information of the target energy storage power station, specifically includes the following steps: Obtain data information from the target energy storage power station; The data information includes data information of the battery cluster, data information of the battery management system, data information of the energy storage converter, and environmental data information of the target energy storage power station; The data information of the battery cluster includes voltage data information, current data information and temperature data information; The data information of the battery management system includes cell voltage data, SOC data, and SOH data; The data information of the energy storage converter includes power data, DC side voltage data, and efficiency data.

3. The collaborative operation and maintenance method for energy storage power stations according to claim 2, characterized in that... Step S2 involves preprocessing the data information obtained in step S1 to construct a panoramic real-time dataset of the target energy storage power station. This process specifically includes the following steps: The data obtained in step S1 is processed using the following formula: In the formula The processed measurement value acquired by the i-th sensor at time t; This represents the raw measurement value acquired by the i-th sensor at time tk; The weighting factor is set. This represents the number of data points within the observation window. This represents the average value of the measurements taken by the i-th sensor within the set observation window. Let be the standard deviation of the measurements taken by the i-th sensor within the set observation window; For device j, construct the corresponding original protocol features. for ,in For equipment type identification, For data frame structure description; set of state parameters Set as , Let L be the measured value of the Lth state variable output by device j; according to and The original bitstream is unpacked into a state vector with physical units. for ; Using the following formula, Convert each component in the equation to a dimensionless standardized value. : In the formula This represents the mean of the corresponding components; This represents the standard deviation of the corresponding component; All Constructing a standardized data descriptor for ; It is used to enable data exchange between various brands and models of sensors.

4. The collaborative operation and maintenance method for energy storage power stations according to claim 3, characterized in that... Step S3, which involves training the digital twin model based on the panoramic real-time dataset obtained in step S2 to obtain a distributed collaborative digital twin model of the target energy storage power station, specifically includes the following steps: Sub-models are established for the battery clusters, battery management system and energy storage converter in the target energy storage system, respectively; Constructing a battery cluster sub-model: The battery cluster model adopts a first-order RC equivalent circuit model; the state equation of the model is expressed as: In the formula Let be the terminal voltage of the battery cluster at time t; Let be the current of the battery cluster at time t; The internal resistance of the battery cluster is ohmic. The polarization internal resistance of the battery cluster; The data sampling interval; It is a time constant; This is the open-circuit voltage of the battery cluster; Constructing a sub-model for the battery management system: A sub-model of the battery management system is constructed using a gated recurrent unit neural network; The forward propagation formula for the model is: In the formula Let be the hidden state at time t; For gated loop unit; Input data for the battery management system sub-model; The SOC state value predicted by the model at time t; The weight vector for SOC prediction; The bias for SOC prediction; The SOH state value predicted by the model at time t; The weight vector predicted by SOH; The bias for SOH prediction; This is the softplus function; Construct a sub-model for the energy storage converter: The energy storage converter sub-model is constructed using the following formula: In the formula AC side power; Let be the efficiency of the energy storage converter corresponding to the kth discrete operating point; This represents the DC-side power corresponding to the kth discrete operating point. DC-side power; For each constructed sub-model, the parameters of the sub-model are updated using the following formula: In the formula The parameter values ​​of sub-model i at time t+1; Let be the parameter values ​​of sub-model i at time t; The set learning rate; Let be the predicted output of sub-model i at time t; This represents the actual output of sub-model i at time t; This is the residual loss function; The gradient of the residual loss function; During training of each sub-model, the following globally consistent loss is constructed: In the formula The global consistency loss value is denoted by n; n is the number of sub-models. The weights of the i-th sub-model are set; For global reference state, , This is the core prediction output of the i-th sub-model at time t; By using global consistency loss, the sub-model can maintain consistency with the overall operating characteristics while fitting its own data.

5. The collaborative operation and maintenance method for energy storage power stations according to claim 4, characterized in that... Step S4, which involves performing distributed collaborative analysis on the target energy storage power station based on the distributed collaborative digital twin model obtained in step S3, and achieving a global operational status assessment of the target energy storage power station, specifically includes the following steps: Constructing a global runtime state vector : In the formula The cumulative charge and discharge capacity of the target energy storage power station. , The total number of battery clusters within the target energy storage power station. This is the current time step for statistics. For battery cluster b at time DC power, The sampling interval; This represents the maximum charge / discharge power in the hourly range. , Indicates the most recent hour. The charging and discharging power of the target energy storage power station; For the current power, the continuous operating time , Let b be the state of charge of the b-th battery cluster at time t. The rated capacity of the b-th battery cluster is... For a given minimum number; The weighted average SOC value. ; Minimum SOH value; The following formula is used as the comprehensive evaluation function: In the formula To comprehensively evaluate the function value; The set credibility weight; The SOC contribution coefficient is set. The contribution coefficient of SOH is set; Based on the obtained comprehensive evaluation function value, the consistency deviation is calculated. for ; Consistency deviation Make a judgment: like If the value exceeds a set threshold, then sub-model i is retrained; simultaneously, adjustments are made. The value is , The set attenuation coefficient.

6. The collaborative operation and maintenance method for energy storage power stations according to claim 5, characterized in that... Step S5, which involves predicting the indicators of the target energy storage power station based on the evaluation results obtained in step S4 using a neural network model and then correcting the prediction results, specifically includes the following steps: Constructing key indicator vectors for ;in For maximum charge and discharge power, Let be the energy conversion efficiency at time t; The input to the neural network model is The output is a future time. Key Indicator Vector Predicted value ; The neural network model employs a three-layer long short-term memory network structure; the LSTM layer updates the memory units. and hidden state The formula for calculating network forward propagation is: In the formula The output of the input gate; It is the sigmoid activation function; Here is the input weight matrix of the input gate; Let be the hidden state weight matrix of the input gate; For the input gate bias; The output of the forget gate; This is the input weight matrix for the forget gate; Let be the hidden state weight matrix of the forget gate; For the offset of the forget gate; This is the output of the output gate; This is the input weight matrix for the output gate; Let be the hidden state weight matrix of the output gate; For the output gate bias; The candidate state of the cell at time t; The input weight matrix represents the candidate cell states; The hidden state weight matrix represents the candidate cell state. This is a bias term for the candidate state of the cell; The cell state at time t; This is element-wise multiplication; Input the hidden state of the last time step into the fully connected output layer to obtain the future time step. Key Indicator Vector Predicted value for ;in To output the weight matrix, For output bias; The residual value was calculated. for ; When the residual value When the value is less than or equal to the set threshold, the output of the sub-model is directly used as the final result; When the residual value When the value exceeds the set threshold, the final result is calculated using the following formula: In the formula This is the final result obtained from the calculation; The set weight value.

7. The collaborative operation and maintenance method for energy storage power stations according to claim 6, characterized in that... Step S6, which involves generating operation and maintenance decision results based on the data information obtained in step S5 to complete the collaborative operation and maintenance of the target energy storage power station, specifically includes the following steps: Based on the data obtained in step S5, an optimization model is constructed to adjust the operation mode of the target energy storage power station; The following formula is used as the objective function of the optimization model: In the formula The objective function value; This is the first weighting coefficient set; The actual charging and discharging power at time k; This is a reference value for charging and discharging power. This is the second weighting coefficient that is set; The energy conversion efficiency of the energy storage power station; This is the set third weighting coefficient; As a risk factor, , The weighting coefficient for health status. The SOH value of the k1th battery cluster is... Temperature weighting coefficient, This represents the highest cell temperature in the battery cluster. The set safe temperature threshold; The following formula is used as the constraint condition for the optimization model: In the formula The charging power of the target energy storage power station at time t; This is the maximum charging power; Let be the discharge power of the target energy storage power station at time t; This represents the maximum discharge power. Minimum SOC value; The maximum SOC value; This is the minimum allowable operating temperature for the cooling system; The actual operating temperature of the cooling system at time t; This is the maximum allowable operating temperature of the cooling system; The constructed optimization model is solved to obtain decision variables, which are then used as the operation and maintenance decision results; the decision variables include , , and the fault status indicator variable of the energy storage power station at time t ; The operation and maintenance decisions are used to complete the collaborative operation and maintenance of the target energy storage power station.

8. The collaborative operation and maintenance method for energy storage power stations according to claim 7, characterized in that... Step S6 further includes the following steps: The instruction execution consistency value is calculated using the following formula: In the formula The instruction execution consistency value for subsystem i; The target value of the instruction for subsystem i; This is the rated reference value of the instruction control quantity corresponding to subsystem i; Determine the consistency of instruction execution values: If the instruction execution consistency value is greater than or equal to the set threshold, the subsystem execution accuracy is determined to meet the requirements. If the instruction execution consistency value is less than the set threshold, it is determined that the subsystem's execution accuracy does not meet the requirements. At this time, the instruction target value is sent to the corresponding subsystem again, and monitoring is performed again.

9. A system for implementing the collaborative operation and maintenance method for an energy storage power station as described in any one of claims 1 to 8, characterized in that... It includes a data acquisition module, a data processing module, a model building module, a status assessment module, a result correction module, and a collaborative operation and maintenance module; the data acquisition module, data processing module, model building module, status assessment module, result correction module, and collaborative operation and maintenance module are connected in series; the data acquisition module is used to acquire data information of the target energy storage power station and upload the data information to the data processing module; The data processing module is used to preprocess the acquired data information based on the received data information in order to construct a panoramic real-time dataset of the target energy storage power station, and upload the data information to the model building module; The model building module is used to train the digital twin model based on the received data and the obtained panoramic real-time dataset to obtain the distributed collaborative digital twin model of the target energy storage power station, and upload the data information to the status assessment module; the status assessment module is used to perform distributed collaborative analysis on the target energy storage power station based on the received data and the obtained distributed collaborative digital twin model, and realize the global operation status assessment of the target energy storage power station, and upload the data information to the result correction module. The result correction module is used to predict the indicators of the target energy storage power station based on the received data and the obtained evaluation results, using a neural network model, and to correct the prediction results and upload the data to the collaborative operation and maintenance module. The collaborative operation and maintenance module is used to generate operation and maintenance decision results based on the received data information and complete the collaborative operation and maintenance of the target energy storage power station.

Images