Energy storage power station risk assessment method and system

Abstract

The present invention relates to the technical field of energy storage power station operation and maintenance, and specifically provides a risk assessment method and system for an energy storage power station. The method obtains an operating status data set of the energy storage power station, where the operating status data set covers multiple equipment monitoring sequences, and each equipment monitoring sequence contains electrical parameter and environmental parameter monitoring data of the energy storage equipment. Risk feature extraction is used to accurately extract an electrical risk feature set and an environmental risk feature set from the operating status data. Based on a dynamic risk assessment model, the extracted risk features are dynamically assessed in real time to generate a comprehensive risk assessment result of the equipment monitoring sequence. Furthermore, an optimized control strategy for the energy storage power station is intelligently generated based on the comprehensive risk assessment result, and the optimized control strategy is fed back to the energy storage power station control system in real time to trigger targeted risk reduction operations, thereby realizing closed-loop management of risk assessment and control strategy, and significantly improving the safe operation level and risk response capability of the energy storage power station.

Description

Technical Field

[0001] The present invention relates to the field of energy storage power station operation and maintenance technology, and in particular to a risk assessment method and system for an energy storage power station. Background Art

[0002] Amid the large-scale application and intelligent development of energy storage power plants, existing technologies for risk assessment still face significant technical bottlenecks and shortcomings. Traditional risk assessment methods often utilize static or quasi-static assessment models, which are unable to reflect the dynamic changes in the operating status of energy storage power plants in real time. During actual operation of energy storage power plants, the electrical and environmental parameters of the equipment are constantly fluctuating. Static assessment models struggle to adapt to this dynamic nature, causing risk assessment results to lag behind actual risk conditions and failing to provide timely and effective decision support for risk mitigation.

[0003] Furthermore, existing technologies often separate risk assessment from control strategy formulation, lacking a closed-loop feedback mechanism. Even if potential risks can be identified, it's difficult to dynamically adjust the energy storage plant's operating parameters based on the risk assessment results to effectively mitigate the risks and optimize the system's control. This disconnect between assessment and control not only reduces the efficiency of energy storage plants but also increases the probability of system failure, posing a serious threat to their safe and stable operation. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, the present invention provides a risk assessment method for an energy storage power station, the method comprising:

[0005] Acquire an operating status data set of the energy storage power station, the operating status data set comprising a plurality of equipment monitoring sequences, each equipment monitoring sequence comprising electrical parameter monitoring data and environmental parameter monitoring data of at least one energy storage device;

[0006] Performing risk feature extraction processing on the operating status data set to obtain an electrical risk feature set and an environmental risk feature set for each equipment monitoring sequence;

[0007] Performing a dynamic risk assessment on the electrical risk feature set and the environmental risk feature set based on a dynamic risk assessment model to generate a comprehensive risk assessment result of the equipment monitoring sequence;

[0008] An optimized control strategy for the energy storage power station is generated based on the comprehensive risk assessment result, and the optimized control strategy is fed back to the energy storage power station control system to trigger risk mitigation operations.

[0009] On the other hand, the present invention also provides a risk assessment system for an energy storage power station, comprising a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0010] Based on the above aspects, the present invention integrates multi-source heterogeneous operating status data to construct a multi-dimensional risk feature system covering electrical parameters and environmental parameters, effectively breaking through the limitations of traditional assessment methods that only focus on single equipment or static indicators. The collaborative analysis of electrical risk features and environmental risk features based on the dynamic risk assessment model not only realizes the real-time risk quantitative assessment of the equipment monitoring sequence, but also reveals the coupling mechanism of potential risk factors through global risk association modeling, significantly improving the accuracy of risk identification. Furthermore, through the closed-loop feedback mechanism of risk assessment results and optimized control strategies, the energy storage power station control system can dynamically adjust the operating parameters according to the real-time risk situation, thereby ensuring the safe and stable operation of the system while achieving the dual goals of risk reduction and energy efficiency optimization, thereby contributing to the intelligent management of the energy storage power station throughout its life cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 It is a schematic diagram of the execution flow of the energy storage power station risk assessment method provided by an embodiment of the present invention.

[0012] Figure 2 Schematic diagram of exemplary hardware and software components of the energy storage power station risk assessment system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 FIG1 is a flow chart of a method for risk assessment of an energy storage power station provided by an embodiment of the present invention. The method for risk assessment of an energy storage power station is introduced in detail below.

[0014] Step S110: Acquire an operating status data set of the energy storage power station, wherein the operating status data set includes a plurality of equipment monitoring sequences, and each equipment monitoring sequence is composed of electrical parameter monitoring data and environmental parameter monitoring data of at least one energy storage device.

[0015] This embodiment uses an energy storage power station as a scenario, in which a large number of energy storage devices of different models are distributed in multiple areas.

[0016] Step S111: The electrical parameter monitoring data and environmental parameter monitoring data of each energy storage device are periodically collected through the sensor network deployed in the energy storage power station. The electrical parameter monitoring data includes voltage, current, charge and discharge power, and battery internal resistance. The environmental parameter monitoring data includes temperature, humidity, air pressure, and device surface vibration intensity.

[0017] In this embodiment, each energy storage device in the energy storage power station is equipped with various sensors. For example, for voltage data, a voltage sensor is provided that collects voltage values ​​every time interval Δt1 (the maximum common sampling interval), forming a voltage data sequence V. A current sensor similarly collects current values ​​every Δt1, forming a current data sequence I. The charge and discharge power is collected by a power sensor at the same Δt1 interval, resulting in a charge and discharge power sequence P. The internal resistance sensor also collects battery internal resistance data at intervals of Δt1, forming a battery internal resistance sequence R. The temperature sensor collects temperature data at intervals of Δt1, forming a temperature sequence T. The humidity sensor, air pressure sensor, and device surface vibration intensity sensor also collect data at intervals of Δt1, generating a humidity sequence H, an air pressure sequence B, and a device surface vibration intensity sequence S, respectively.

[0018] Step S112: performing data cleaning processing on the collected electrical parameter monitoring data and the environmental parameter monitoring data to obtain a standardized monitoring data set, wherein the data cleaning processing includes outlier removal, time stamp alignment and data format unification.

[0019] In this embodiment, considering that the collected data may have various problems, it needs to be cleaned.

[0020] Step S1121: calling the corresponding electrical parameter and environmental parameter normal range library according to the model identifier of the energy storage device, performing a first anomaly detection process based on the device model on the electrical parameter monitoring data and the environmental parameter monitoring data, and eliminating abnormal data points that exceed the parameter range corresponding to the device model.

[0021] For example, energy storage device X is Model-A, and its normal voltage range is V_min_A to V_max_A. If the collected voltage value V_a is outside this normal voltage range, the data point is determined to be an outlier and must be removed from the voltage data sequence V. Regarding current, the normal current range for Model-A devices is I_min_A to I_max_A. If the collected current value I_a is outside this range, the data point is similarly removed. Parameters such as charge and discharge power, battery internal resistance, temperature, humidity, air pressure, and device surface vibration intensity all undergo similar outlier detection and removal based on the normal ranges corresponding to their respective device models.

[0022] Step S1122: performing a second anomaly detection process on the environmental parameter monitoring data, removing abnormal data points that exceed a preset environmental parameter range, and performing spatial interpolation filling on missing data points based on adjacent sensor data.

[0023] In addition to device model-based anomaly detection, there are also preset universal ranges for environmental parameters (environmental thresholds for indoor and outdoor devices). For example, the preset normal range for temperature is T_min to T_max. If the collected temperature value T_b is outside this range, it is removed from the temperature sequence T. If data is missing at a location in the temperature sequence T, but sensors at adjacent locations collect temperature values ​​T_c and T_d, spatial interpolation methods such as linear interpolation and kriging interpolation can be used. If linear interpolation is used, the temperature value at the missing location is calculated based on T_c and T_d and added to the temperature sequence T. Environmental parameters such as humidity, air pressure, and device surface vibration intensity also undergo anomalous data points that fall outside the preset ranges and missing data points are filled in using spatial interpolation.

[0024] Step S1123: Time synchronization processing is performed on the processed electrical parameter monitoring data and the environmental parameter monitoring data to ensure that the timestamps of each data point are aligned according to a unified time reference.

[0025] Because data collected by different sensors may have slight discrepancies, the timestamps of all data need to be unified. Using the power plant's standard clock as the unified time reference, the timestamps of processed electrical and environmental parameter monitoring data are adjusted. For example, if the timestamp of a voltage data point deviates from the standard clock, it is adjusted to align with the standard clock, ensuring consistency across all data in the time dimension.

[0026] Step S1124: converting the time-synchronized data into a standard data format, and classifying and storing the data according to the device identifier to generate the standardized monitoring data set.

[0027] For example, voltage data can be uniformly converted to a set standard format, such as a numerical format with a fixed number of decimal places. Then, based on the identifiers of individual energy storage devices, all data belonging to the same device can be grouped together for storage. For example, all processed data from device X can be stored in a set data structure, ultimately forming a standardized monitoring data set.

[0028] Step S113: Divide the standardized monitoring data set into a plurality of equipment monitoring sequences according to a preset time window, and store the equipment monitoring sequences in a distributed database.

[0029] Assuming a preset time window of Δt2, starting from the starting point of the standardized monitoring data set, data of each Δt2 duration is divided into a device monitoring sequence. For example, the first device monitoring sequence contains all electrical and environmental parameter data from the starting point to the starting point + Δt2. Each device monitoring sequence has a unique ID, which can be used to accurately locate the corresponding sequence in the distributed database. The distributed database consists of multiple storage nodes, and different device monitoring sequences may be stored on different nodes, enabling distributed data storage and management.

[0030] Step S120: performing risk feature extraction processing on the operating status data set to obtain an electrical risk feature set and an environmental risk feature set for each equipment monitoring sequence.

[0031] In this embodiment, after obtaining multiple equipment monitoring sequences, risk feature extraction is performed for each equipment monitoring sequence.

[0032] Step S121: Performing time series fluctuation analysis on the electrical parameter monitoring data in the equipment monitoring sequence to obtain the operating stability characteristics and energy conversion abnormality characteristics of the energy storage device, wherein the operating stability characteristics include voltage fluctuation rate and current balance, and the energy conversion abnormality characteristics include charge and discharge efficiency deviation value and power mutation frequency.

[0033] Taking one of the equipment monitoring sequences as an example, its electrical parameter monitoring data is analyzed.

[0034] For example, step S1211: calculating the sliding standard deviation of the voltage sequence in the electrical parameter monitoring data according to a preset time window to generate a voltage fluctuation rate.

[0035] For example, set a preset time window Δt3 (Δt3 is less than Δt2) and slide the window length Δt3 over the voltage sequence V, starting from the beginning of the sequence. For each voltage value within the Δt3 window, calculate its sliding standard deviation. For example, within the nth Δt3 window, the voltage values ​​are V_n1, V_n2, …, V_nm (m is determined by Δt3 and Δt1). By calculating the sliding standard deviation of these values, we obtain the voltage fluctuation rate V_vol_n corresponding to that window. The voltage fluctuation rates corresponding to all windows are combined to form the voltage fluctuation rate sequence V_vol.

[0036] Step S1212: calculating the root mean square difference of each phase current at the same moment for the multi-phase current sequence in the electrical parameter monitoring data to generate a current balance.

[0037] In this embodiment, it is assumed that the energy storage device has multiphase currents, denoted as I_1, I_2, …, I_k (where k is the number of phases). At the same time t, the RMS difference of each phase current, I_1_t, I_2_t, …, I_k_t, is calculated. The RMS current difference at that time, I_eq_t, is obtained using conventional calculation methods in the relevant art. The RMS current differences at each time are combined to form a current balance sequence, I_eq.

[0038] Step S1213: calculating the absolute value of the sliding average deviation between the actual efficiency and the rated efficiency for the charge and discharge power sequence in the electrical parameter monitoring data, and generating a charge and discharge efficiency deviation value.

[0039] For example, given the rated efficiency as Eff_rated, a time window Δt4 (Δt4 less than Δt2) is set on the charge and discharge power sequence P. The actual efficiency Eff is calculated for the power values ​​within each Δt4 window. For example, within the pth Δt4 window, the actual efficiency Eff_p is calculated from the power values. Then, the sliding average of |Eff_p - Eff_rated| is calculated to obtain the charge and discharge efficiency deviation value Eff_dev_p corresponding to that window. The charge and discharge efficiency deviation values ​​corresponding to all windows are combined to form the charge and discharge efficiency deviation value sequence Eff_dev.

[0040] Step S1214: performing mutation point detection on the power sequence in the electrical parameter monitoring data, counting the frequency of exceeding the dynamic mutation threshold within a unit time window, and generating a power mutation frequency.

[0041] For example, assume a unit time window Δt5 (Δt5 less than Δt2) and a dynamic mutation threshold Th. A sliding window of Δt5 is run over the power sequence P, detecting power mutation points within each window. If the power value within the window changes by more than Th, it is considered a mutation. The number of mutations within each Δt5 window is counted to obtain the power mutation frequency sequence P_mut_num.

[0042] Step S1215: obtaining operation stability characteristics and energy conversion abnormality characteristics according to the voltage fluctuation rate, current balance, charge and discharge efficiency deviation value and power mutation frequency.

[0043] In this embodiment, the voltage fluctuation rate sequence V_vol and the current balance sequence I_eq are combined to form the operation stability feature, and the charge and discharge efficiency deviation value sequence Eff_dev and the power mutation frequency sequence P_mut_num are combined to form the energy conversion abnormality feature.

[0044] Step S122: Perform spatial correlation analysis on the environmental parameter monitoring data in the equipment monitoring sequence to obtain the environmental impact characteristics and regional risk diffusion characteristics of the energy storage equipment, wherein the environmental impact characteristics include the temperature gradient correlation value and the humidity distribution difference, and the regional risk diffusion characteristics include the coupling degree of environmental parameters of adjacent equipment and the strength of the risk propagation path.

[0045] The environmental parameter monitoring data of the same equipment monitoring sequence is taken as an example for analysis.

[0046] For example, step S1221: for the temperature sequence in the environmental parameter monitoring data, calculate the sliding average of the absolute values ​​of the temperature gradients of adjacent sensors at the same time to generate a temperature gradient correlation value.

[0047] Consider the temperature values ​​of adjacent sensors in a temperature sequence T. At the same time t, let the temperature values ​​of adjacent sensors be T_t1 and T_t2. Calculate the absolute value of their temperature gradients, |T_t1 - T_t2|. Calculate the sliding average of the absolute values ​​of the temperature gradients within each window, using a sliding window Δt6 (Δt6 is less than Δt2). For example, within the qth Δt6 window, calculate the sliding average of the absolute values ​​of the temperature gradients at each time point within the window to obtain the temperature gradient correlation value T_grad_q corresponding to that window. Combine the temperature gradient correlation values ​​corresponding to all windows to form the temperature gradient correlation value sequence T_grad.

[0048] Step S1222: For the humidity sequence in the environmental parameter monitoring data, calculate the coefficient of dispersion of the humidity of each sensor at the same time to generate the humidity distribution difference.

[0049] In the humidity sequence H, at the same time t, calculate the coefficient of dispersion of the humidity values ​​H_t1, H_t2, …, H_tl (l is the number of sensors) for each sensor. Using conventional calculation methods from related technologies, we obtain the humidity coefficient of dispersion at that moment, i.e., the humidity distribution difference H_diff_t. The humidity distribution differences at each moment are combined to form the humidity distribution difference sequence H_diff.

[0050] Step S1223: For the temperature and humidity sequences in the environmental parameter monitoring data, the maximum value of the sliding window cross-correlation function of the adjacent device sensor data is calculated to generate the adjacent device environmental parameter coupling degree.

[0051] For the temperature sequence T and humidity sequence H, consider the sensor data of adjacent devices. Within each sliding window Δt7 (Δt7 is less than Δt2), calculate the cross-correlation function of the temperature and humidity sensor data of adjacent devices. For example, within the rth Δt7 window, calculate the cross-correlation function of the temperature and humidity sensor data of adjacent devices, and take the maximum value as the adjacent device environmental parameter coupling degree Coup_r corresponding to that window. Combine the adjacent device environmental parameter coupling degrees corresponding to all windows to form the adjacent device environmental parameter coupling degree sequence Coup.

[0052] Step S1224: For the temperature gradient correlation value sequence in the environmental parameter monitoring data, combined with the energy storage power station topology, calculate the cumulative gradient change rate within the sliding window of the temperature propagation path between each device to generate the risk propagation path strength.

[0053] Given the topology of an energy storage power station, a sliding window Δt8 (Δt8 less than Δt2) is used as the interval on the temperature gradient correlation value sequence T_grad. For each window, the cumulative gradient change rate of the temperature propagation path between each device is calculated based on the topology. For example, in the s-th Δt8 window, the temperature propagation path is determined based on the topology, and the cumulative gradient change rate along this path is calculated to obtain the risk propagation path strength Risk_path_s corresponding to this window. The risk propagation path strengths corresponding to all windows are combined to form the risk propagation path strength sequence Risk_path.

[0054] Step S1225: Obtain environmental impact characteristics and regional risk diffusion characteristics based on the temperature gradient correlation value, humidity distribution difference, adjacent equipment environmental parameter coupling degree, and risk propagation path strength.

[0055] The temperature gradient correlation value sequence T_grad and the humidity distribution difference sequence H_diff are combined to form the environmental impact feature. The adjacent equipment environmental parameter coupling degree sequence Coup and the risk propagation path strength sequence Risk_path are combined to form the regional risk diffusion feature.

[0056] Step S123: After uniformly performing Z-Score standardization processing on the operation stability characteristics, energy conversion abnormality characteristics, environmental impact characteristics and regional risk diffusion characteristics, weighted fusion and dynamic coupling are performed according to the preset electrical characteristic weight coefficient and environmental characteristic weight coefficient to generate the electrical risk characteristic set and environmental risk characteristic set of the equipment monitoring sequence, wherein the electrical risk characteristic set includes the weighted fusion result of the operation stability characteristics and the energy conversion abnormality characteristics, and the environmental risk characteristic set includes the dynamic coupling result of the environmental impact characteristics and the regional risk diffusion characteristics.

[0057] For example, step S1231: subtract the mean of each characteristic dimension from the voltage fluctuation rate, current balance, charge and discharge efficiency deviation value, and power mutation frequency, and divide by the standard deviation to generate a first standardized vector.

[0058] For the voltage fluctuation sequence V_vol, calculate its mean μ_V_vol and standard deviation σ_V_vol. For each value V_vol_i in the sequence, calculate the standardized voltage fluctuation value using the formula (V_vol_i - μ_V_vol) / σ_V_vol. Perform the same operation for the current balance sequence I_eq, the charge-discharge efficiency deviation sequence Eff_dev, and the power mutation frequency sequence P_mut_num to obtain the standardized current balance value, charge-discharge efficiency deviation value, and power mutation frequency value, respectively. Combine these standardized values ​​into the first standardized vector N1.

[0059] Step S1232: Subtract the mean of each characteristic dimension from the temperature gradient correlation value, humidity distribution difference, adjacent equipment environmental parameter coupling degree, and risk propagation path strength, and divide by the standard deviation to generate a second standardized vector.

[0060] For the temperature gradient correlation value sequence T_grad, calculate its mean μ_T_grad and standard deviation σ_T_grad. For each value T_grad_i in the sequence, calculate the standardized temperature gradient correlation value using the formula (T_grad_i - μ_T_grad) / σ_T_grad. Perform the same operation for the humidity distribution difference sequence H_diff, the adjacent device environmental parameter coupling sequence Coup, and the risk propagation path strength sequence Risk_path, respectively, to obtain the standardized humidity distribution difference value, adjacent device environmental parameter coupling value, and risk propagation path strength value. Combine these standardized values ​​into the second standardized vector N2.

[0061] Step S1233: After mapping the first normalized vector and the second normalized vector to the same numerical interval, linearly superimpose the voltage fluctuation rate and the current balance degree in the first normalized vector according to a preset first weight ratio to generate a weighted value of operation stability.

[0062] Assume that the first normalized vector N1 and the second normalized vector N2 are mapped to the interval [0, 1]. Assume that the preset first weight ratio is w1_1 and w1_2 (w1_1 + w1_2 = 1). For the normalized voltage fluctuation rate value V_vol_norm and the normalized current balance value I_eq_norm in the first normalized vector N1, the operational stability weighted value Stab_weight is calculated according to the formula V_vol_norm*w1_1+I_eq_norm*w1_2.

[0063] Step S1234: linearly superimpose the charge and discharge efficiency deviation value and the power mutation frequency in the first normalized vector according to a preset second weight ratio to generate an energy conversion abnormality weighted value.

[0064] Assume that the preset second weight ratio is w2_1 and w2_2 (w2_1+w2_2=1). For the standardized charge and discharge efficiency deviation value Eff_dev_norm and the standardized power mutation frequency value P_mut_num_norm in the first standardized vector N1, the energy conversion abnormality weighted value Energy_abn_weight is calculated according to the formula Eff_dev_norm*w2_1+P_mut_num_norm*w2_2.

[0065] Step S1235: linearly superimpose the temperature gradient correlation value and the humidity distribution difference in the second normalized vector according to a preset third weight ratio to generate an environmental impact weighted value.

[0066] Assume that the preset third weight ratio is w3_1 and w3_2 (w3_1+w3_2=1). For the standardized temperature gradient correlation value T_grad_norm and the standardized humidity distribution difference value H_diff_norm in the second standardized vector N2, the environmental impact weighted value Env_impact_weight is calculated according to the formula T_grad_norm*w3_1+H_diff_norm*w3_2.

[0067] Step S1236: Linearly superimpose the coupling degree of the adjacent equipment environmental parameters in the second standardized vector and the risk propagation path strength according to the preset fourth weight ratio to generate a regional risk diffusion weighted value; wherein, the first weight ratio, the second weight ratio, the third weight ratio and the fourth weight ratio are dynamically adjusted according to the contribution of each feature in the historical fault data.

[0068] Assuming the fourth preset weight ratio is w4_1 and w4_2 (w4_1 + w4_2 = 1), the regional risk diffusion weighted value Risk_diff_weight is calculated using the formula Coup_norm * w4_1 + Risk_path_norm * w4_2 for the standardized adjacent equipment environmental parameter coupling value Coup_norm and the standardized risk propagation path strength value Risk_path_norm in the second standardized vector N2. Here, w1_1, w1_2, w2_1, w2_2, w3_1, w3_2, w4_1, and w4_2 are dynamically adjusted based on the contribution of each feature to the fault impact in the historical fault data.

[0069] Step S1237: performing nonlinear function mapping on the weighted value of the operating stability and the weighted value of the energy conversion anomaly according to a preset electrical fusion coefficient to generate the electrical risk feature set.

[0070] Assume that the preset electrical fusion coefficient is α, and the operation stability weighted value Stab_weight and the energy conversion abnormality weighted value Energy_abn_weight are calculated through a nonlinear function f (such as some activation function) according to the formula f(Stab_weight*α+Energy_abn_weight*(1-α)) to obtain the elements in the electrical risk feature set, and combine these elements to form the electrical risk feature set Electrical_risk_set.

[0071] Step S1238: Perform spatiotemporal convolution operation on the weighted value of the environmental impact and the weighted value of the regional risk diffusion according to the preset environmental fusion coefficient to generate the environmental risk feature set; wherein, the electrical fusion coefficient and the environmental fusion coefficient are automatically optimized through the back propagation process of the dynamic risk assessment model.

[0072] In this embodiment, the aforementioned embodiment has obtained an environmental impact weighted value set recorded as Env_impact_weight_set, which includes the corresponding values ​​of the environmental impact weighted value Env_impact_weight at each time and space position, and a regional risk diffusion weighted value set recorded as Risk_diff_weight_set, which includes the corresponding values ​​of the regional risk diffusion weighted value Risk_diff_weight at each time and space position, and the preset environmental fusion coefficient is β.

[0073] Determine a convolution kernel Conv_kernel that adapts to the spatiotemporal dimensions of the data. Its size and shape are determined by the spatiotemporal characteristics presented by Env_impact_weight_set and Risk_diff_weight_set. For example, if the data is sampled at fixed intervals in time and distributed in two dimensions in space, the convolution kernel may be a three-dimensional structure, with two dimensions corresponding to space and one dimension corresponding to time.

[0074] For each spatiotemporal position point in Env_impact_weight_set and Risk_diff_weight_set (let the time dimension be t and the spatial dimension be x, y in two dimensions), let Conv_kernel cover a specific spatiotemporal neighborhood with that point as the center. Within this neighborhood, multiply the values ​​of the corresponding positions of Env_impact_weight_set and Risk_diff_weight_set, then multiply them by the value of the corresponding position of Conv_kernel, and then accumulate these products.

[0075] For example, at the spatiotemporal position point (t, x, y), in the neighborhood covered by Conv_kernel, the corresponding value of Env_impact_weight_set is E(t+i, x+j, y+k), the corresponding value of Risk_diff_weight_set is R(t+i, x+j, y+k), and the corresponding value of Conv_kernel is K(i, j, k), where -n<=i<=n, -m<=j<=m, -p<=k<=p (n, m, p are determined according to the size of Conv_kernel), then the calculation result of this point is ∑(E(t+i, x+j, y+k)*R(t+i, x+j, y+k)*K(i, j, k)). This calculation is performed on all spatiotemporal position points to obtain a new intermediate set Intermediate_set.

[0076] Afterwards, each value in Intermediate_set is multiplied by β to finally generate the environmental risk feature set Environmental_risk_set.

[0077] During dynamic risk assessment model training, the electrical fusion coefficient α and the environmental fusion coefficient β are automatically optimized through a backpropagation process. During model training, the predicted risk assessment results are compared with the sample risk assessment labels, and adjustments to α and β are calculated based on the difference. α and β are adjusted according to predefined rules, such as the proportion of the adjustment, to bring the predicted results closer to the sample risk assessment labels.

[0078] Step S130: performing a dynamic risk assessment on the electrical risk feature set and the environmental risk feature set based on a dynamic risk assessment model to generate a comprehensive risk assessment result of the equipment monitoring sequence.

[0079] After obtaining the electrical risk feature set Electrical_risk_set and the environmental risk feature set Environmental_risk_set, a dynamic risk assessment model is used to perform dynamic risk assessment.

[0080] Step S131: Call the electrical risk assessment subnetwork in the dynamic risk assessment model to perform adaptive weight allocation processing on the electrical risk feature set to generate a first risk level score for the energy storage device, wherein the adaptive weight allocation processing includes adjusting the contribution weight of the operating stability feature in the first risk level score according to the dynamic change trend of the voltage fluctuation rate and the current balance.

[0081] The electrical risk feature set Electrical_risk_set is input into the electrical risk assessment subnetwork. The operational stability features in the electrical risk feature set are composed of the voltage fluctuation rate sequence V_vol and the current balance sequence I_eq.

[0082] To determine the contribution of operational stability characteristics to the first-level risk assessment, we analyze the dynamic trends of V_vol and I_eq. For example, we observe the changes in V_vol over multiple consecutive time windows (assuming a window length of Δt3) and calculate the rate of change within each window: (V_vol(t+Δt3)-V_vol(t)) / V_vol(t). This yields the V_vol rate sequence V_vol_rate. Similarly, we calculate the rate of change of I_eq within each window to obtain I_eq_rate.

[0083] Next, analyze the relationship between V_vol_rate and I_eq_rate, for example, by calculating their correlation. If the two are positively correlated, this means that as V_vol increases, I_eq also tends to increase. Based on this correlation, adjust the weights of V_vol and I_eq in the operational stability feature. If changes in V_vol_rate have a more significant impact on risk, appropriately increase the weight of V_vol in the operational stability feature and correspondingly decrease the weight of I_eq.

[0084] By continuously adjusting the weights based on the dynamic trends of these two factors, the contribution weight of the operational stability feature to the first risk level score is determined. The weighted operational stability feature is then compared with the energy conversion anomaly features in the electrical risk feature set (including the charge and discharge efficiency deviation value sequence Eff_dev and the power mutation frequency sequence P_mut_num). Following the calculation rules set by the electrical risk assessment subnetwork, for example, by multiplying each feature value by its corresponding weight and then adding them together, the First_risk_score, the first risk level score for the energy storage device, is generated.

[0085] For example, in another embodiment, step S131 may include:

[0086] Step S1311: calculating a first covariance matrix between the dynamic change trend of the voltage fluctuation rate and the dynamic change trend of the current balance degree, and determining a dynamic contribution weight coefficient of the operation stability feature based on the eigenvalue distribution of the first covariance matrix.

[0087] Taking the voltage fluctuation rate sequence V_vol and the current balance degree sequence I_eq as examples, to calculate the covariance matrix between their dynamic change trends, we first determine an analysis window length Δt. Within this window, for the voltage fluctuation rate, the difference between adjacent time points is calculated to obtain the voltage fluctuation rate change sequence ΔV_vol, i.e., ΔV_vol(t)=V_vol(t)-V_vol(t-1). Similarly, for the current balance degree, the difference between adjacent time points is calculated to obtain the current balance degree change sequence ΔI_eq, i.e., ΔI_eq(t)=I_eq(t)-I_eq(t-1).

[0088] Then, for each time point t, the product of ΔV_vol(t) and ΔI_eq(t) is calculated. These products are summed over the entire analysis window and then divided by the number of data points in the window minus 1 to obtain the covariance value Cov(ΔV_vol, ΔI_eq). The covariance values ​​calculated for different analysis windows are combined to form the first covariance matrix Cov_matrix.

[0089] Next, the first covariance matrix Cov_matrix is ​​subjected to eigenvalue decomposition to obtain its eigenvalues. The dynamic contribution weight coefficient of the operation stability feature is determined based on the distribution of these eigenvalues. For example, if a certain eigenvalue is large, it means that the correlation between the change trends of the two in the corresponding direction is strong, then the weight of the voltage fluctuation rate or current balance associated with the eigenvalue in the operation stability feature will be adjusted accordingly. The specific adjustment method is to divide each eigenvalue by the sum of all eigenvalues ​​to obtain the proportion of each eigenvalue. These proportions are used as the basis for adjusting the dynamic contribution weight coefficients of the voltage fluctuation rate and current balance in the operation stability feature.

[0090] Step S1312: performing exponential smoothing on the charge and discharge efficiency deviation value to generate a first smoothed sequence, performing sliding window statistics on the power mutation frequency to generate a first frequency distribution histogram, and determining the dynamic contribution weight coefficient of the energy conversion abnormality feature based on the information entropy ratio of the first smoothed sequence and the first frequency distribution histogram.

[0091] Perform exponential smoothing on the charge-discharge efficiency deviation sequence Eff_dev. Assume the smoothing coefficient is α (0 < α < 1), and the initial smoothing value is S_0 = Eff_dev(0). For subsequent time points t, the smoothing value is calculated according to the formula S_t = α * Eff_dev(t) + (1 - α) * S_t - 1, thus generating the first smoothed sequence S.

[0092] For the power mutation frequency sequence P_mut_num, a sliding window length Δt_window is set. Within each sliding window, the distribution of power mutation frequencies is statistically analyzed to generate a first frequency distribution histogram. The horizontal axis of the histogram represents the range of power mutation frequencies, and the vertical axis represents the number of times a frequency occurs within this range.

[0093] Next, calculate the information entropy (Entropy_S) of the first smoothed sequence S and the information entropy (Entropy_hist) of the first frequency distribution histogram within the smoothing window. Entropy is calculated by summing the probability of each value multiplied by the negative of its logarithm. The ratio of these two entropies, Entropy_S / Entropy_hist, is used to determine the dynamic contribution weights of the charge-discharge efficiency deviation and power mutation frequency in the energy conversion anomaly signature. For example, a larger ratio indicates a more regular change in the charge-discharge efficiency deviation, and its weight in the energy conversion anomaly signature will increase accordingly.

[0094] Step S1313: normalizing the dynamic contribution weight coefficient corresponding to the operation stability feature and the dynamic contribution weight coefficient corresponding to the energy conversion abnormality feature to generate a first weight distribution vector.

[0095] The dynamic contribution weight coefficients of the voltage fluctuation rate and current balance in the operation stability feature are combined and recorded as Weight_run_stab. The dynamic contribution weight coefficients of the charge and discharge efficiency deviation value and power mutation frequency in the energy conversion abnormality feature are combined and recorded as Weight_energy_abn.

[0096] Normalize Weight_run_stab and Weight_energy_abn. First, add all weight coefficients in Weight_run_stab and Weight_energy_abn to obtain the sum Sum_weights. Then, divide each weight coefficient in Weight_run_stab and Weight_energy_abn by Sum_weights to obtain the normalized weight coefficients. Arrange these normalized weight coefficients in a specific order to generate the first weight allocation vector Weight_vector_1.

[0097] Step S1314: performing a weighted summation calculation on the standardized operation stability characteristics and energy conversion abnormality characteristics according to the first weight distribution vector to generate a first risk level score for the energy storage device, wherein the input feature dimension of the weighted summation calculation matches the dimension of the first weight distribution vector.

[0098] The standardized operation stability characteristics are composed of the standardized voltage fluctuation rate V_vol_norm and the standardized current balance I_eq_norm. The energy conversion abnormality characteristics are composed of the standardized charge and discharge efficiency deviation value Eff_dev_norm and the standardized power mutation frequency P_mut_num_norm.

[0099] These standardized eigenvalues ​​are weighted and summed with the first weight distribution vector Weight_vector_1. That is, the first risk level score First_risk_score = V_vol_norm*Weight_vector_1(1)+I_eq_norm*Weight_vector_1(2)+Eff_dev_norm*Weight_vector_1(3)+P_mut_num_norm*Weight_vector_1(4), where Weight_vector_1(1), Weight_vector_1(2), Weight_vector_1(3), and Weight_vector_1(4) are the weight coefficients of the corresponding positions in the first weight distribution vector Weight_vector_1. In this way, the first risk level score of the energy storage device is generated.

[0100] Step S132: Call the environmental risk assessment subnetwork in the dynamic risk assessment model to perform spatiotemporal correlation mapping processing on the environmental risk feature set to generate a second risk level score for the energy storage device, wherein the spatiotemporal correlation mapping processing includes adjusting the influence coefficient of the regional risk diffusion characteristic in the second risk level score according to the spatiotemporal distribution characteristics of the temperature gradient correlation value and the risk propagation path intensity.

[0101] The environmental risk feature set Environmental_risk_set is input into the environmental risk assessment subnetwork. The regional risk diffusion features in the environmental risk feature set are composed of the temperature gradient correlation value sequence T_grad and the risk propagation path intensity sequence Risk_path.

[0102] Analyze the spatiotemporal distribution characteristics of T_grad and Risk_path. For T_grad, observe its changes at different spatial locations (corresponding to different sensor locations) and at different time points. For example, if T_grad values ​​in certain areas are high and continue to rise within a certain time period, this indicates a significant temperature gradient change in that area. For Risk_path, analyze its changes over time across different devices. If the Risk_path value between a certain device gradually increases, it indicates that the risk propagation path is strengthening.

[0103] Adjust the influence coefficient of regional risk diffusion characteristics in the second risk level score based on the spatiotemporal distribution characteristics of T_grad and Risk_path. For example, if a region has a high T_grad value and a high Risk_path value, indicating a high likelihood of risk spread in that region, then appropriately increase the influence coefficient of regional risk diffusion characteristics in the second risk level score.

[0104] In specific operations, an adjustment rule is determined by analyzing the temporal and spatial distribution patterns of T_grad and Risk_path. For example, a function is set that takes the average, maximum, and rate of change of T_grad and Risk_path within a specific temporal and spatial region as input and outputs an adjustment factor. Based on this adjustment factor, the initial impact coefficient of the regional risk diffusion characteristics in the second risk level score is adjusted.

[0105] After adjusting the impact coefficient, the regional risk diffusion characteristics and the environmental impact characteristics in the environmental risk feature set (including the humidity distribution difference sequence H_diff and the adjacent equipment environmental parameter coupling degree sequence Coup) are combined according to the calculation rules set by the environmental risk assessment subnetwork. For example, each characteristic value is multiplied by its corresponding adjusted weight and then accumulated to generate the second risk level score of the energy storage equipment, Second_risk_score.

[0106] For example, in another embodiment, step S132 may include:

[0107] Step S1321: Perform spatial interpolation processing on the temperature gradient correlation value to generate a temperature distribution surface, perform time sliding average processing on the intensity of the risk propagation path to generate an intensity change curve, and calculate the spatiotemporal influence coefficient of the regional risk diffusion characteristics based on the spatiotemporal superposition result of the temperature distribution surface and the intensity change curve.

[0108] For the temperature gradient correlation value sequence T_grad, assume that these temperature gradient correlation values ​​are collected by sensors at different spatial locations at different time points. When performing spatial interpolation processing, a two-dimensional coordinate system is constructed with spatial position as the horizontal axis and time as the vertical axis. At each time point, based on the temperature gradient correlation value at the known spatial location, an appropriate interpolation method (such as linear interpolation, spline interpolation, or kriging interpolation) is used to calculate the temperature gradient correlation values ​​at other spatial locations, thereby generating the temperature distribution surface T_surface.

[0109] For the risk propagation path strength sequence, Risk_path, a sliding window of length Δt_window_time is set. Within each sliding window, the risk propagation path strength values ​​are averaged to obtain the average strength value within the window. Over time, these average strength values ​​form the strength variation curve, Risk_path_curve.

[0110] The temperature distribution surface T_surface and the intensity variation curve Risk_path_curve are superimposed in time and space. For example, at each time point and spatial location, the corresponding value on the temperature distribution surface is multiplied by the value on the intensity variation curve at the corresponding time point. These products are then summed over the entire time and space range. According to a certain normalization rule (such as dividing by the total number of points in the time and space range), the time and space influence coefficient Space_time_coeff of the regional risk diffusion characteristics is obtained.

[0111] Step S1322: Gaussian filtering is performed on the humidity distribution difference to generate a second smoothed sequence, cross-correlation analysis is performed on the coupling degrees of the adjacent device environmental parameters to generate a coupling degree correlation matrix, and the dynamic influence coefficient of the environmental impact characteristic is determined based on the variance of the second smoothed sequence and the spectral radius of the coupling degree correlation matrix.

[0112] Gaussian filtering is performed on the humidity distribution difference sequence H_diff. A Gaussian kernel is set, with its size and standard deviation determined based on the characteristics of the data. On the humidity distribution difference sequence, with each data point as the center, the Gaussian kernel is overlaid on the adjacent data points. The value at the corresponding position of the Gaussian kernel is multiplied by the value of the data point, and these products are added together to obtain the smoothed value at that position after filtering. This operation is repeated for the entire sequence to generate the second smoothed sequence H_smooth.

[0113] A cross-correlation analysis is performed on the adjacent device environmental parameter coupling sequence Coup. A matrix, the coupling correlation matrix Coup_matrix, is constructed for the adjacent device environmental parameter coupling values ​​between different device pairs. The rows and columns of the matrix correspond to different devices, and the matrix elements represent the adjacent device environmental parameter coupling values ​​between the corresponding device pairs.

[0114] Calculate the variance Var_H_smooth of the second smoothed sequence H_smooth and the spectral radius Spectral_radius_Coup_matrix of the coupling matrix Coup_matrix. The spectral radius is the maximum modulus of the matrix's eigenvalues. Based on these two values, determine the dynamic influence coefficient of the environmental impact feature. For example, a function can be set that takes the variance and spectral radius as input and outputs an adjustment factor, which serves as the dynamic influence coefficient Dynamic_influence_coeff of the environmental impact feature.

[0115] Step S1323: normalizing the spatiotemporal influence coefficient and the dynamic influence coefficient to generate a second weight distribution vector.

[0116] Combine the spatial-temporal influence coefficient (Space_time_coeff) of the regional risk diffusion feature and the dynamic influence coefficient (Dynamic_influence_coeff) of the environmental impact feature. First, calculate the sum of these two coefficients (Sum_coeffs). Then, divide both the spatial-temporal influence coefficient (Space_time_coeff) and the dynamic influence coefficient (Dynamic_influence_coeff) by Sum_coeffs to obtain normalized coefficients. Arrange these normalized coefficients in a specific order to generate the second weight distribution vector (Weight_vector_2).

[0117] Step S1324: performing a weighted summation calculation on the standardized regional risk diffusion characteristics and the environmental impact characteristics according to the second weight distribution vector to generate a second risk level score for the energy storage device, wherein the input feature dimension of the weighted summation calculation matches the dimension of the second weight distribution vector.

[0118] The standardized regional risk diffusion characteristics are composed of the standardized temperature gradient correlation value T_grad_norm, the standardized risk propagation path intensity Risk_path_norm, and the standardized coupling degree of adjacent equipment environmental parameters Coup_norm. The environmental impact characteristics are composed of the standardized humidity distribution difference H_diff_norm.

[0119] These standardized eigenvalues ​​are weighted and summed with the second weight distribution vector Weight_vector_2. That is, the second risk level score Second_risk_score = T_grad_norm*Weight_vector_2(1)+Risk_path_norm*Weight_vector_2(2)+Coup_norm*Weight_vector_2(3)+H_diff_norm*Weight_vector_2(4), where Weight_vector_2(1), Weight_vector_2(2), Weight_vector_2(3), and Weight_vector_2(4) are the weight coefficients of the corresponding positions in the second weight distribution vector Weight_vector_2. In this way, the second risk level score of the energy storage device is generated.

[0120] Step S133: Normalize the first risk level score and the second risk level score to a preset unified interval respectively, perform nonlinear superposition processing on the normalized scores based on the risk coupling module in the dynamic risk assessment model, generate a comprehensive risk index for the energy storage device, and determine the comprehensive risk assessment result of the equipment monitoring sequence based on the comparison result of the comprehensive risk index and the preset risk threshold.

[0121] First, a preset uniform interval is determined, assuming it is the interval [0, 1]. For the first risk level score First_risk_score, a normalization method is used. For example, the maximum value Max_First and the minimum value Min_First in First_risk_score are found. For each score value Score_First, the formula (Score_First-Min_First) / (Max_First-Min_First) is calculated and normalized to the interval [0, 1] to obtain the normalized first risk level score Normalized_First.

[0122] Similarly, for the second risk level score Second_risk_score, find its maximum value Max_Second and minimum value Min_Second, calculate each score value Score_Second according to the formula (Score_Second-Min_Second) / (Max_Second-Min_Second), normalize it to the interval [0, 1], and obtain the normalized second risk level score Normalized_Second.

[0123] Normalized_First and Normalized_Second are input into the risk coupling module in the dynamic risk assessment model. The risk coupling module superimposes the two using a nonlinear function. Assuming that the nonlinear function is f(x, y), the composite risk index Composite_risk_index = f(Normalized_First, Normalized_Second).

[0124] A risk threshold, Risk_threshold, is preset and Composite_risk_index is compared with Risk_threshold. If Composite_risk_index is greater than Risk_threshold, the comprehensive risk assessment result of the equipment monitoring sequence is determined to be high risk; if Composite_risk_index is less than or equal to Risk_threshold, the comprehensive risk assessment result is determined to be low risk or medium risk (the specific risk can be determined based on further subdivision rules).

[0125] Step S140: generating an optimized control strategy for the energy storage power station based on the comprehensive risk assessment result, and feeding the optimized control strategy back to the energy storage power station control system to trigger risk mitigation operations.

[0126] Step S141: Determine the risk device set and risk propagation path set of the energy storage power station based on the comprehensive risk assessment result, wherein the risk device set includes energy storage devices whose comprehensive risk index exceeds a preset risk threshold, and the risk propagation path set includes the environmental parameter coupling degree and electrical parameter correlation degree between each energy storage device in the risk device set.

[0127] Based on the comprehensive risk assessment results obtained above, the composite risk index Composite_risk_index corresponding to each energy storage device is compared with the preset risk threshold Risk_threshold. If the Composite_risk_index of a certain energy storage device is greater than the Risk_threshold, the energy storage device is included in the risk device set Risk_device_set.

[0128] For each energy storage device in the risk_device_set, analyze the degree of coupling between their environmental parameters and the degree of correlation between their electrical parameters to determine the risk propagation path set, risk_path_set. For example, for two energy storage devices A and B, calculate the degree of coupling between their environmental parameters, such as the temperature gradient correlation value and humidity distribution difference, as well as the degree of correlation between their electrical parameters, such as voltage and current. These parameter values ​​are then incorporated into the portion of the risk_path_set related to A and B.

[0129] Step S142: Prioritize the risk equipment set to generate an equipment control priority list, wherein the priority sorting includes: calculating a dynamic priority coefficient according to a preset weight allocation rule based on the comprehensive risk index, the strength of the risk propagation path, and the position weight of the energy storage equipment in the power station topology, and generating an equipment control priority list based on the dynamic priority coefficient.

[0130] For each device in the risk device set Risk_device_set, first determine its composite risk index Composite_risk_index, risk propagation path strength (obtain the risk propagation path strength value Risk_path_strength related to the device from Risk_path_set), and the location weight Location_weight of the device in the power station topology (determined according to the power station topology, for example, the location weight of the device close to the key node is higher).

[0131] A weight distribution rule is preset, assuming that the weight of the composite risk index is w1, the weight of the risk propagation path strength is w2, and the weight of the location weight is w3, and w1 + w2 + w3 = 1. For each risky device, calculate the dynamic priority coefficient Dynamic_priority_coefficient = w1*Composite_risk_index+w2*Risk_path_strength+w3*Location_weight.

[0132] Sort the devices in Risk_device_set according to the calculated Dynamic_priority_coefficient and generate a device control priority list Device_priority_list from high to low.

[0133] Step S143: Generate an optimized control strategy for the energy storage power station based on the equipment control priority list and the risk propagation path set, wherein the optimized control strategy includes a charge and discharge power adjustment plan for each risky device, an environmental control parameter optimization plan, and an equipment isolation switching plan.

[0134] Based on the device control priority list Device_priority_list and the risk propagation path set Risk_path_set, an optimized control strategy is formulated for each risky device.

[0135] For high-priority risk equipment, first develop a charge and discharge power adjustment plan. For example, if a high overall risk index for a risky device is due to abnormal charge and discharge power, analyze the power anomaly based on its electrical parameter monitoring data, such as the charge and discharge efficiency deviation value (Eff_dev) and the power mutation frequency (P_mut_num). If Eff_dev is large, indicating that the charge and discharge efficiency deviates significantly from the rated efficiency, it may be necessary to reduce the charging power or adjust the discharge power curve to optimize the charging and discharging process.

[0136] At the same time, develop an environmental control parameter optimization plan. Obtain the degree of environmental parameter coupling between the device and surrounding devices from the risk_path_set, such as the temperature gradient correlation value T_grad and the humidity distribution difference H_diff. If T_grad is large, it indicates a significant temperature difference between the device and surrounding devices. Environmental control parameters such as the air conditioning cooling capacity or ventilation volume in the area where the device is located may need to be adjusted to reduce the temperature gradient and minimize the impact of the environment on the device.

[0137] For high-risk devices that may affect other equipment, develop device isolation and switching plans. Based on the plant topology and the risk propagation path information in risk_path_set, if a device's risk is found to be likely to spread rapidly to other critical equipment, consider electrically isolating the device or switching to a backup device to prevent the risk from spreading.

[0138] The charging and discharging power adjustment plan, environmental control parameter optimization plan, and equipment isolation switching plan for each risky device are integrated to form the optimization control strategy Optimization_control_strategy of the energy storage power station.

[0139] Step S144: Feedback the optimization control strategy to the energy storage power station control system to trigger risk reduction operations.

[0140] Step S1441: adjusting the operating power parameters of the risk equipment according to the charge-discharge power adjustment scheme, and monitoring the changing trends of the adjusted operating stability characteristics and the energy conversion abnormality characteristics in real time.

[0141] The charge and discharge power adjustment plan in the optimization control strategy is sent to the energy storage power station control system. The control system adjusts the operating power parameters of the risk equipment according to the plan, such as changing the charging current or discharge power setting.

[0142] During the adjustment process, monitor the operating stability characteristics of the risky equipment (including voltage fluctuation rate V_vol and current balance I_eq) and energy conversion anomaly characteristics (including charge and discharge efficiency deviation value Eff_dev and power mutation frequency P_mut_num) in real time. For example, observe whether V_vol gradually decreases after power adjustment, indicating that voltage fluctuations are stabilizing; whether I_eq is closer to the ideal equilibrium state; whether Eff_dev gradually decreases and approaches the rated efficiency; and whether P_mut_num decreases, indicating that power mutation conditions are improving.

[0143] Step S1442: adjusting the environmental control equipment in the area where the risk equipment is located according to the environmental control parameter optimization plan, and monitoring the dynamic response data of the adjusted environmental impact characteristics and the regional risk diffusion characteristics in real time.

[0144] Based on the environmental control parameter optimization plan in the optimization control strategy, the environmental control equipment in the area where the risk equipment is located is adjusted. For example, if the plan requires increasing the cooling capacity of the air conditioner, the control system sends a command to the air conditioner to increase its cooling capacity.

[0145] Simultaneously, dynamic response data for the region's environmental impact characteristics (including temperature gradient correlation value T_grad and humidity distribution difference H_diff) and regional risk diffusion characteristics (including the coupling degree Coup of adjacent equipment environmental parameters and the intensity of the risk propagation path Risk_path) are monitored in real time. For example, monitoring is performed to see whether T_grad decreases with increasing cooling capacity, whether H_diff remains within a reasonable range, and whether Coup decreases, indicating a weakening of the environmental parameter coupling between adjacent equipment. A decrease in Risk_path indicates a weakening of the risk propagation path.

[0146] Step S1443: Electrically isolate the risky equipment or perform backup equipment switching operations according to the equipment isolation and switching plan, and re-input the adjusted operating stability characteristics and energy conversion abnormality characteristics into the dynamic risk assessment model for incremental risk assessment. If the decline in the updated comprehensive risk index does not reach the preset threshold or the fluctuation of the continuous assessment results exceeds the preset range, the generation of the secondary optimization control strategy is terminated, and the current optimization control strategy is directly output.

[0147] According to the equipment isolation and switching plan in the optimized control strategy, electrically isolate the risky equipment or switch to backup equipment. If electrical isolation is performed, the electrical connection between the risky equipment and other equipment is cut off; if backup equipment is switched, the load is transferred from the risky equipment to the backup equipment.

[0148] After the operation is completed, the adjusted operational stability characteristics (V_vol, I_eq) and energy conversion anomaly characteristics (Eff_dev, P_mut_num) of the device are re-entered into the dynamic risk assessment model for incremental risk assessment. The model recalculates the composite risk index (Updated_composite_risk_index) based on these updated characteristic values.

[0149] A reduction threshold (Reduction_threshold) and a fluctuation threshold (Fluctuation_threshold) are preset. If the reduction in the updated composite risk index relative to the previous composite risk index does not reach the reduction threshold, or if the fluctuation of multiple consecutive evaluation results exceeds the fluctuation threshold, the current optimization control strategy is considered to have achieved satisfactory results or further optimization may introduce instability. The generation of the secondary optimization control strategy is terminated, and the current optimization control strategy, optimization_control_strategy, is directly output.

[0150] For example, the method further includes:

[0151] Step S151: Mapping the comprehensive risk assessment results to a three-dimensional topology of the energy storage power station according to a preset segmentation threshold, distinguishing different risk level intervals through discrete color identifiers, and dynamically associating risk mitigation operation suggestions corresponding to each interval.

[0152] First, determine the preset segmentation threshold, for example, divide the comprehensive risk index range into multiple intervals, such as [0, 0.3], (0.3, 0.6], (0.6, 1]. For the comprehensive risk index Composite_risk_index corresponding to each comprehensive risk assessment result, determine the interval to which it belongs.

[0153] The 3D topology of the energy storage power station is displayed, and risky devices in different intervals are marked with different discrete color identifiers. For example, risky devices in the interval [0, 0.3] are represented in green, those in the interval (0.3, 0.6) are represented in yellow, and those in the interval (0.6, 1) are represented in red.

[0154] For each risk level interval, corresponding risk mitigation action suggestions are associated. For example, for the green interval, regular inspections are recommended; for the yellow interval, increased monitoring frequency is recommended; and for the red interval, immediate execution of relevant actions in the optimization and control strategy is recommended.

[0155] Step S152: Generate a risk reduction operation suggestion list according to the optimization control strategy, and dynamically associate the operation suggestion list with the risk devices in the three-dimensional topology map for display.

[0156] Based on the optimization control strategy, Operation_suggestion_list is generated. This list includes suggestions for adjusting the charge and discharge power, optimizing environmental control parameters, and isolating and switching equipment for each risky device.

[0157] In the 3D topology map, dynamically associate the operation suggestions in the Operation_suggestion_list with the corresponding risky devices. For example, when you click a risky device in the topology map, a pop-up window displays the operation suggestions for that device.

[0158] Step S153: updating the comprehensive risk assessment results and the execution effect of the optimized control strategy in real time, and displaying the risk index change trend and the effectiveness status of the control strategy in a visual interface.

[0159] During system operation, the latest comprehensive risk assessment results, namely the updated comprehensive risk index Updated_composite_risk_index, and the execution effect data of the optimized control strategy are obtained in real time, such as whether the charging and discharging power is adjusted in place, whether the environmental control parameters reach the set values, and whether the equipment isolation switching is successful.

[0160] In the visualization interface, a risk index trend chart is plotted, showing how the Updated_composite_risk_index changes over time. With time as the horizontal axis and the composite risk index as the vertical axis, a point is drawn at each new Updated_composite_risk_index value, and adjacent points are connected with lines to form a risk index change curve. This allows operators to intuitively see how the risk index changes over time and as optimization and control strategies are implemented.

[0161] The effectiveness of the optimization and control strategy is also displayed. For the charge and discharge power adjustment plan, if the operating power parameters of the risk equipment are successfully adjusted to the set value, the "Charge and discharge power adjustment successful" icon will be displayed next to the corresponding equipment icon on the visualization interface. If an abnormality occurs during the adjustment process, such as the power cannot reach the set value, the prompt message "Charge and discharge power adjustment abnormality, please check" will be displayed.

[0162] For the environmental control parameter optimization solution, if the environmental control device adjusts the parameters as set, for example, the air conditioner successfully increases or decreases the cooling capacity, and the environmental parameters of the area (such as temperature and humidity) begin to change in the expected direction, the message "Environmental control parameter optimization is effective" is displayed. If the environmental control device fails to respond to the command correctly, or the environmental parameters do not change as expected, the message "Environmental control parameter optimization failed, need to be investigated" is displayed.

[0163] For the equipment isolation and switching plan, if the electrical isolation or backup equipment switching operation is successfully completed, the relevant display area of ​​the device on the visual interface will highlight the status information that the device has been isolated or switched to the backup equipment; if a fault occurs during the switching process, such as the electrical connection cannot be cut off normally or the backup equipment cannot be started, "Equipment isolation switching failure, emergency processing" will be displayed.

[0164] By updating comprehensive risk assessment results and displaying the execution effects of optimized control strategies in real time, operators can promptly understand the risk status of the energy storage power station and the effectiveness of control measures to make further decisions.

[0165] In a possible implementation, the above embodiment may further include step S210: a step of training the dynamic risk assessment model. For details, please refer to the following embodiment.

[0166] Step S211: Acquire a sample operation status data set and a corresponding sample risk assessment label, wherein the sample risk assessment label includes equipment failure records, risk event occurrence time, and risk impact range.

[0167] Sample operating status data sets are collected from historical data records. These data sets have a similar format to the previously acquired energy storage power station operating status data sets and also contain multiple equipment monitoring sequences. Each equipment monitoring sequence consists of electrical parameter monitoring data and environmental parameter monitoring data of the energy storage equipment.

[0168] At the same time, we collect sample risk assessment labels corresponding to the sample operating status data set. For example, for a certain equipment monitoring sequence, we record whether the energy storage device has ever failed. If a failure has occurred, we record the failure type, start time, and end time in detail as the equipment failure record. We also record the specific time when the risk event occurred. We also assess the scope of the risk event's impact on surrounding equipment or the entire power plant operation, such as which areas of equipment are affected and how many other energy storage devices are involved. This constitutes the sample risk assessment label.

[0169] Step S212: performing feature extraction processing on the sample operating status data set to obtain a sample electrical risk feature set and a sample environmental risk feature set.

[0170] This step is similar to the risk feature extraction process performed on the operating status data set in the previous step S120. For each device monitoring sequence in the sample operating status data set, the electrical parameter monitoring data and the environmental parameter monitoring data are analyzed respectively.

[0171] For the electrical parameter monitoring data, time series fluctuation analysis is performed according to the method in step S121. For example, the sliding standard deviation of the voltage series is calculated to obtain the voltage fluctuation rate, the root mean square difference of the multi-phase current series is calculated to obtain the current balance, and the absolute value of the sliding mean deviation between the actual efficiency and the rated efficiency of the charge and discharge power series is calculated to obtain the charge and discharge efficiency deviation value. The power series is subjected to mutation point detection to calculate the power mutation frequency, thereby obtaining the operating stability characteristics and energy conversion abnormality characteristics.

[0172] For environmental parameter monitoring data, refer to step S122 to perform spatial correlation analysis. For example, the temperature sequence calculates the sliding average of the absolute values ​​of the temperature gradients of adjacent sensors at the same time to obtain the temperature gradient correlation value. The humidity sequence calculates the coefficient of dispersion of the humidity of each sensor at the same time to obtain the humidity distribution difference. The temperature and humidity sequences calculate the maximum value of the sliding window cross-correlation function of the sensor data of adjacent devices to obtain the coupling degree of the environmental parameters of adjacent devices. The temperature gradient correlation value sequence, combined with the power station topology, calculates the cumulative gradient change rate within the sliding window of the temperature propagation path between each device to obtain the risk propagation path strength, thereby obtaining the environmental impact characteristics and regional risk diffusion characteristics.

[0173] Then, these features are Z-Score standardized according to step S123, and weighted fusion and dynamic coupling are performed according to the preset electrical feature weight coefficients and environmental feature weight coefficients to generate a sample electrical risk feature set and a sample environmental risk feature set.

[0174] Step S213: constructing an initial risk assessment model, wherein the initial risk assessment model includes an electrical risk assessment subnetwork, an environmental risk assessment subnetwork, and a risk coupling module.

[0175] The construction of the electrical risk assessment subnetwork must consider adaptive weighting of the set of electrical risk features. For example, the network structure should be designed to accept inputs such as voltage fluctuation rate, current balance, charge / discharge efficiency deviation, and power mutation frequency. The network can contain multiple layers of neurons, connected by weights. These weights are initially randomly set but are subsequently adjusted during training.

[0176] The environmental risk assessment subnetwork must be able to map the spatial and temporal correlations of the environmental risk feature set. For example, the network structure must be able to process features such as temperature gradient correlations, humidity distribution differences, coupling between environmental parameters of adjacent devices, and the strength of risk propagation paths. It may include structures capable of processing spatiotemporal data, such as convolutional or recurrent layers, to capture the temporal and spatial correlations of environmental features.

[0177] The risk coupling module is used to perform nonlinear superposition processing on the outputs of the electrical risk assessment subnetwork and the environmental risk assessment subnetwork. This module can be a simple nonlinear function module that receives the outputs from both subnetworks and calculates them using a predefined nonlinear function to output a comprehensive risk index. These three components are combined to form the initial risk assessment model.

[0178] Step S214: inputting the sample electrical risk feature set and the sample environmental risk feature set into the initial risk assessment model to obtain a predicted risk assessment result.

[0179] The sample electrical risk feature set and the sample environmental risk feature set are input into the electrical risk assessment subnetwork and the environmental risk assessment subnetwork of the initial risk assessment model respectively.

[0180] The electrical risk assessment subnetwork processes the sample electrical risk feature set and, based on its internally defined weights and calculation rules, adaptively weights features such as voltage fluctuation and current balance, outputting a first predicted risk level score. For example, neurons within the electrical risk assessment subnetwork perform weighted summation and other calculations on the input features. After processing through an activation function, the resulting output value serves as the first predicted risk level score.

[0181] The environmental risk assessment subnetwork processes the sample environmental risk feature set and, using its spatiotemporal correlation mapping structure, analyzes features such as temperature gradient correlation values ​​and risk transmission path strength to output a second predicted risk level score. For example, a convolutional layer extracts features from spatiotemporal data, followed by a fully connected layer and other operations to produce the output value, the second predicted risk level score.

[0182] The risk coupling module receives the first predicted risk level score and the second predicted risk level score, performs superposition processing according to the set nonlinear function, and outputs the predicted risk assessment result, that is, the predicted comprehensive risk index.

[0183] Step S215: Calculate the model loss based on the difference between the predicted risk assessment result and the sample risk assessment label, and update the parameters of the initial risk assessment model based on the gradient descent algorithm until the model converges to obtain the trained dynamic risk assessment model.

[0184] The model loss is measured by calculating the difference between the predicted risk assessment result (the predicted comprehensive risk index) and the sample risk assessment label. For example, a loss function such as the mean squared error loss function can be used to calculate the average squared error between the predicted comprehensive risk index and the actual risk level reflected in the sample risk assessment label.

[0185] The parameters of the initial risk assessment model are updated using a gradient descent algorithm. This algorithm determines the direction of parameter updates by calculating the gradient of the loss function with respect to model parameters (such as the weights between neurons in the electrical and environmental risk assessment subnetworks). For example, for a weight parameter w, the update formula is w = w - learning_rate * grad_w, where learning_rate is the pre-set learning rate and grad_w is the gradient of the loss function with respect to w.

[0186] This process—inputting sample data, calculating losses, and updating parameters—is repeated until the model converges. Model convergence means the loss function no longer decreases significantly, for example, when the change in the loss function over multiple consecutive iterations is less than a minimal threshold. At this point, a trained dynamic risk assessment model is obtained, which can more accurately assess the risks of energy storage power plants.

[0187] Step S220: The sample risk assessment label is generated through the following steps.

[0188] Step S221: performing event correlation analysis on each device monitoring sequence in the sample operating status data set to determine whether a failure event or a risk warning event occurs in the energy storage device within a monitoring time period.

[0189] For each device monitoring sequence in the sample operating status data set, analyze in detail how the electrical parameter monitoring data and environmental parameter monitoring data change over time. For example, check whether the voltage sequence fluctuates beyond the normal range, whether the current sequence is unbalanced, whether there are abnormal changes in charge and discharge power, and whether environmental parameters such as temperature and humidity deviate from the normal range.

[0190] By analyzing these data, it is possible to determine whether the energy storage device has experienced any failure events during the monitoring period, such as whether the battery is damaged, resulting in abnormal charging and discharging, or whether a risk warning event has been triggered, such as the temperature is too high and close to the dangerous threshold but has not yet caused a failure.

[0191] Step S222: If a fault event occurs, a first type of risk assessment label is generated based on the fault type, fault duration, and repair cost.

[0192] If a failure event is determined for the energy storage device corresponding to a device monitoring sequence, the type of failure, such as a battery short circuit or line aging failure, should be recorded in detail. The duration from the occurrence of the failure to repair should also be recorded, as well as the cost of repairing the failure, including the cost of replacing parts and labor costs.

[0193] Based on this information, a first-class risk assessment label is generated. For example, the fault type is coded, with different codes representing different types of faults. The fault duration is quantified according to predefined rules, such as dividing it into several time periods, with each period corresponding to a quantified value. The repair cost is also quantified according to a predefined ratio. These quantified values ​​are then combined according to a predefined format to form a first-class risk assessment label that comprehensively reflects the severity and impact of the fault event.

[0194] Step S223: If a risk warning event occurs but no fault is triggered, a second type of risk assessment label is generated based on the warning level, risk diffusion speed, and control response effect.

[0195] When a risk warning event is determined for an energy storage device corresponding to a device monitoring sequence but no failure has occurred, the warning level is first determined. This may be determined by pre-established warning rules, such as the degree to which parameters such as temperature and pressure deviate from their normal range. The speed of risk diffusion is assessed, for example, by analyzing changes in environmental parameters and correlations with electrical parameters of adjacent devices to determine the spatial and temporal spread of the risk. Furthermore, the effectiveness of regulatory measures implemented in response to the risk warning event is recorded, such as whether the relevant parameters have returned to their normal range after the adjustments.

[0196] A second-type risk assessment label is generated based on the warning level, risk diffusion rate, and regulatory response effect. For example, the warning level is quantified, the risk diffusion rate is numerically quantified according to set standards, and the regulatory response effect is also quantified using set indicators. These quantified values ​​are then combined according to a set format to form a second-type risk assessment label to reflect the potential threat level of the risk warning event.

[0197] Step S224: If no risk event occurs, the equipment operation stability index is segmented and quantified according to the preset safety threshold, and combined with the statistical distribution characteristics of the environmental parameter fluctuation range, a third type of risk assessment label consistent with the fault event label dimension is generated.

[0198] For equipment monitoring sequences without any failure events or risk warnings, analyze equipment operational stability indicators, such as voltage fluctuation and current balance, and quantify these indicators in segments according to pre-set safety thresholds. For example, different voltage fluctuation ranges can be set, with each corresponding to a quantified value.

[0199] At the same time, the fluctuation range of environmental parameters (such as temperature and humidity) is statistically analyzed, and their distribution characteristics, such as mean and variance, are analyzed. Quantitative processing is performed based on these statistical characteristics.

[0200] The quantified values ​​of the equipment operational stability index and the quantified values ​​of the environmental parameter fluctuation range are combined according to a set format to generate a third-category risk assessment label. This third-category risk assessment label has the same dimension as the first and second-category risk assessment labels, allowing for unified processing later.

[0201] Step S225: performing normalized coding processing on the first-category risk assessment label, the second-category risk assessment label, and the third-category risk assessment label to generate the sample risk assessment label in a unified format.

[0202] Normalize the first, second, and third category risk assessment labels to a uniform range. For example, for each quantized value in each label, find its corresponding maximum and minimum values. Then normalize each quantized value using the formula (value-min) / (max-min), where value is the original quantized value, and min and max are the minimum and maximum values ​​of the feature corresponding to that quantized value, respectively.

[0203] After normalization, the first, second, and third risk assessment labels are encoded to a unified format. For example, the normalized quantized values ​​of the different types of labels are arranged in a set order to form a unified vector form, which serves as the final sample risk assessment label for supervised learning in subsequent model training.

[0204] Figure 2 A schematic diagram illustrating exemplary hardware and software components of an energy storage power plant risk assessment system 100 that can implement the present invention is provided in some embodiments of the present invention. For example, a processor 120 can be used in the energy storage power plant risk assessment system 100 to perform the functions of the present invention.

[0205] Energy storage power plant risk assessment system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the energy storage power plant risk assessment method of the present invention. Although only one server is shown in this invention, for convenience, the functions described in this invention can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0206] For example, the energy storage power station risk assessment system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the energy storage power station risk assessment system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present invention may be implemented based on these program instructions. The energy storage power station risk assessment system 100 also includes an I / O interface 150 between the computer and other input and output devices.

[0207] For ease of explanation, only one processor is described in the energy storage power station risk assessment system 100. However, it should be noted that the energy storage power station risk assessment system 100 in the present invention may also include multiple processors, so the steps performed by one processor described in the present invention may also be performed jointly or individually by multiple processors. For example, if the processor of the energy storage power station risk assessment system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually in one processor. For example, the first processor performs step A and the second processor performs step B, or the first processor and the second processor perform steps A and B together.

[0208] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When a processor executes the computer-executable instructions, the above energy storage power station risk assessment method is implemented.

[0209] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A risk assessment method for an energy storage power station, characterized in that: The method comprises: Acquire an operating status data set of the energy storage power station, the operating status data set comprising multiple equipment monitoring sequences, each equipment monitoring sequence comprising electrical parameter monitoring data and environmental parameter monitoring data of at least one energy storage device; wherein the electrical parameter monitoring data comprises voltage, current, charge / discharge power, and battery internal resistance; and the environmental parameter monitoring data comprises temperature, humidity, air pressure, and equipment surface vibration intensity; Perform risk feature extraction processing on the operating status data set to obtain an electrical risk feature set and an environmental risk feature set for each device monitoring sequence; specifically, the process includes: performing time series fluctuation analysis processing on the electrical parameter monitoring data in the device monitoring sequence to obtain the operating stability characteristics and energy conversion abnormality characteristics of the energy storage device, wherein the operating stability characteristics include the voltage fluctuation rate and the current balance, and the energy conversion abnormality characteristics include the charge and discharge efficiency deviation value and the power mutation frequency; performing spatial correlation analysis processing on the environmental parameter monitoring data in the device monitoring sequence to obtain the environmental impact characteristics and regional risk diffusion characteristics of the energy storage device, wherein the environmental impact characteristics include the temperature gradient correlation value and the humidity distribution difference, and the regional risk diffusion characteristics include the coupling degree of the environmental parameters of adjacent devices and the strength of the risk propagation path; Performing a dynamic risk assessment on the electrical risk feature set and the environmental risk feature set based on a dynamic risk assessment model to generate a comprehensive risk assessment result of the equipment monitoring sequence; An optimized control strategy for the energy storage power station is generated based on the comprehensive risk assessment result, and the optimized control strategy is fed back to the energy storage power station control system to trigger risk mitigation operations.

2. The energy storage power station risk assessment method according to claim 1, characterized in that: The risk feature extraction process is performed on the operating status data set to obtain an electrical risk feature set and an environmental risk feature set for each equipment monitoring sequence, further comprising: After uniformly performing Z-Score standardization processing on the operation stability characteristics, energy conversion abnormality characteristics, environmental impact characteristics and regional risk diffusion characteristics, weighted fusion and dynamic coupling are performed according to the preset electrical characteristic weight coefficient and environmental characteristic weight coefficient to generate the electrical risk characteristic set and environmental risk characteristic set of the equipment monitoring sequence, wherein the electrical risk characteristic set includes the weighted fusion result of the operation stability characteristics and the energy conversion abnormality characteristics, and the environmental risk characteristic set includes the dynamic coupling result of the environmental impact characteristics and the regional risk diffusion characteristics.

3. The energy storage power station risk assessment method according to claim 2, characterized in that: The dynamic risk assessment model is used to dynamically assess the electrical risk feature set and the environmental risk feature set to generate a comprehensive risk assessment result for the equipment monitoring sequence, including: Invoking an electrical risk assessment subnetwork in the dynamic risk assessment model to perform adaptive weight allocation processing on the electrical risk feature set to generate a first risk level score for the energy storage device, wherein the adaptive weight allocation processing includes adjusting a contribution weight of the operating stability feature in the first risk level score based on dynamic change trends of the voltage fluctuation rate and the current balance degree; Calling the environmental risk assessment subnetwork in the dynamic risk assessment model to perform spatiotemporal correlation mapping processing on the environmental risk feature set to generate a second risk level score for the energy storage device, wherein the spatiotemporal correlation mapping processing includes adjusting an influence coefficient of the regional risk diffusion feature in the second risk level score based on the spatiotemporal distribution characteristics of the temperature gradient correlation value and the risk propagation path intensity; The first risk level score and the second risk level score are respectively normalized to a preset unified interval, and the normalized scores are subjected to nonlinear superposition processing based on the risk coupling module in the dynamic risk assessment model to generate a comprehensive risk index of the energy storage device. The comprehensive risk assessment result of the equipment monitoring sequence is determined based on the comparison result of the comprehensive risk index with the preset risk threshold.

4. The energy storage power station risk assessment method according to claim 2, characterized in that: Generating an optimized control strategy for the energy storage power station according to the comprehensive risk assessment results includes: Determining a risk device set and a risk propagation path set of the energy storage power station based on the comprehensive risk assessment result, wherein the risk device set includes energy storage devices whose comprehensive risk index exceeds a preset risk threshold, and the risk propagation path set includes the environmental parameter coupling degree and electrical parameter correlation degree between each energy storage device in the risk device set; Prioritizing the risky equipment set to generate an equipment control priority list, wherein the prioritization process includes: calculating a dynamic priority coefficient according to a preset weight allocation rule based on a comprehensive risk index, the strength of the risk propagation path, and the position weight of the energy storage equipment in the power station topology, and generating the equipment control priority list based on the dynamic priority coefficient; An optimized control strategy for the energy storage power station is generated based on the equipment control priority list and the risk propagation path set, wherein the optimized control strategy includes a charge and discharge power adjustment scheme for each risky device, an environmental control parameter optimization scheme, and an equipment isolation switching scheme.

5. The energy storage power station risk assessment method according to claim 4, characterized in that: Feeding back the optimized control strategy to the energy storage power station control system to trigger risk mitigation operations includes: Adjusting the operating power parameters of the risk equipment according to the charge and discharge power adjustment plan, and monitoring the changing trends of the adjusted operating stability characteristics and the energy conversion abnormality characteristics in real time; Adjusting the environmental control equipment in the area where the risk equipment is located according to the environmental control parameter optimization plan, and monitoring the dynamic response data of the adjusted environmental impact characteristics and the regional risk diffusion characteristics in real time; According to the equipment isolation and switching plan, the risk equipment is electrically isolated or the backup equipment is switched, and the adjusted operation stability characteristics and energy conversion abnormality characteristics are re-input into the dynamic risk assessment model for incremental risk assessment. If the decline in the updated comprehensive risk index does not reach the preset threshold or the fluctuation of the continuous assessment results exceeds the preset range, the generation of the secondary optimization control strategy is terminated and the current optimization control strategy is directly output.

6. The energy storage power station risk assessment method according to claim 1, characterized in that: The step of obtaining the operating status data set of the energy storage power station includes: The sensor network deployed in the energy storage power station periodically collects electrical parameter monitoring data and environmental parameter monitoring data of each energy storage device; Performing data cleaning on the collected electrical parameter monitoring data and the collected environmental parameter monitoring data to obtain a standardized monitoring data set, wherein the data cleaning includes outlier removal, timestamp alignment, and data format unification; The standardized monitoring data set is divided into a plurality of equipment monitoring sequences according to a preset time window, and the equipment monitoring sequences are stored in a distributed database.

7. The energy storage power station risk assessment method according to claim 6, characterized in that: The data cleaning process is performed on the collected electrical parameter monitoring data and the collected environmental parameter monitoring data to obtain a standardized monitoring data set, including: Calling the corresponding electrical parameter and environmental parameter normal range library according to the model identifier of the energy storage device, performing a first anomaly detection process based on the device model on the electrical parameter monitoring data and the environmental parameter monitoring data, and eliminating abnormal data points that exceed the parameter range corresponding to the device model; Performing a second anomaly detection process on the environmental parameter monitoring data, removing abnormal data points that exceed a preset environmental parameter range, and performing spatial interpolation filling on missing data points based on adjacent sensor data; Performing time synchronization processing on the processed electrical parameter monitoring data and the environmental parameter monitoring data to ensure that the timestamps of each data point are aligned according to a unified time reference; The time-synchronized data is converted into a standard data format and classified and stored according to device identifiers to generate the standardized monitoring data set.

8. The energy storage power station risk assessment method according to claim 1, characterized in that: The dynamic risk assessment model is trained by the following steps: Obtaining a sample operating status data set and a corresponding sample risk assessment label, wherein the sample risk assessment label includes equipment failure records, risk event occurrence time, and risk impact scope; Performing feature extraction processing on the sample operating status data set to obtain a sample electrical risk feature set and a sample environmental risk feature set; Constructing an initial risk assessment model, wherein the initial risk assessment model includes an electrical risk assessment subnetwork, an environmental risk assessment subnetwork, and a risk coupling module; Inputting the sample electrical risk feature set and the sample environmental risk feature set into the initial risk assessment model to obtain a predicted risk assessment result; The model loss is calculated based on the difference between the predicted risk assessment result and the sample risk assessment label, and the parameters of the initial risk assessment model are updated based on the gradient descent algorithm until the model converges to obtain the trained dynamic risk assessment model.

9. The energy storage power station risk assessment method according to claim 8, characterized in that: The sample risk assessment label is generated by the following steps: Performing event correlation analysis on each device monitoring sequence in the sample operating status data set to determine whether a failure event or a risk warning event occurs in the energy storage device within a monitoring time period; If a failure event occurs, a first-class risk assessment label is generated based on the failure type, failure duration, and repair cost; If a risk warning event occurs but no fault is triggered, a second-class risk assessment label is generated based on the warning level, risk diffusion speed, and regulatory response effect; If no risk event occurs, the equipment operation stability index is segmented and quantified according to the preset safety threshold, and combined with the statistical distribution characteristics of the environmental parameter fluctuation range, a third-category risk assessment label with the same dimension as the fault event label is generated; The first type of risk assessment label, the second type of risk assessment label, and the third type of risk assessment label are normalized and encoded to generate the sample risk assessment labels in a unified format.

10. A risk assessment system for an energy storage power station, characterized in that: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the energy storage power station risk assessment method described in any one of claims 1 to 9 above.

Images