Operation and Maintenance Management Method and System for Hydropower Station

Abstract

The present invention relates to the technical field of operation and maintenance management of hydropower stations, and specifically provides an operation and maintenance management method and system for hydropower stations, aiming to improve operation and maintenance efficiency and equipment reliability. First, a real-time monitoring data set of the operating equipment of the hydropower station is obtained, covering the sensor data streams of multiple monitoring time windows. Then, dynamic feature extraction operations are performed to generate a state feature set containing features related to equipment operation stability, environmental coupling, and abnormal fluctuations. Next, using a preset abnormal detection strategy network, the abnormal state matching process is carried out on the state feature set to determine the abnormal level and the parameter of the influence range. Finally, according to these parameters, an equipment operation and maintenance optimization strategy set is generated, including an equipment control instruction sequence and a maintenance resource scheduling plan, and is fed back to the hydropower station control center to trigger operation and maintenance operations, realizing the intelligence and precision of the operation and maintenance management of the hydropower station and ensuring the safe and stable operation of the hydropower station.

Description

Technical Field

[0001] The present invention relates to the technical field of operation and maintenance management of hydropower stations, and in particular, to an operation and maintenance management method and system for hydropower stations. Background Art

[0002] In the field of operation and maintenance management of hydropower stations, with the continuous expansion of the scale of hydropower stations and the increasing improvement of the degree of automation, higher requirements are put forward for the real-time monitoring of the operating status of equipment and precise operation and maintenance. However, the existing operation and maintenance management methods for hydropower stations still face many challenges, which limit the further improvement of operation and maintenance efficiency and equipment reliability.

[0003] Traditional operation and maintenance management methods often rely on regular inspections and manual experience judgments, and it is difficult to achieve continuous and comprehensive monitoring of the operating status of equipment. Even if some hydropower stations introduce sensor technology to collect equipment operation data, it is mostly limited to simple threshold comparison or trend analysis, and it is impossible to deeply explore the complex features and potential laws behind the data. In addition, the existing methods have obvious deficiencies in anomaly detection, and it is difficult to accurately identify the types, levels and influence scopes of abnormal states, resulting in a lack of scientific basis for operation and maintenance decisions, unreasonable allocation of operation and maintenance resources, and even possible delays in the timing of fault handling, posing a threat to the safe and stable operation of hydropower stations.

[0004] More critically, the existing operation and maintenance management methods lack an intelligent and automated operation and maintenance strategy generation and feedback mechanism. Even if equipment anomalies can be detected, it often requires manual intervention for strategy formulation and resource scheduling, with slow response speed and low operation and maintenance efficiency. Summary of the Invention

[0005] In view of the above-mentioned problems, in combination with the first aspect of the present invention, the present invention provides an operation and maintenance management method for hydropower stations, and the method includes:

[0006] Obtain a real-time monitoring data set of the operating equipment of the hydropower station, and the real-time monitoring data set includes sensor data streams of multiple monitoring time windows;

[0007] Perform a dynamic feature extraction operation on the real-time monitoring data set to generate a state feature set of the operating equipment of the hydropower station, and the state feature set includes equipment operation stability features, environmental coupling features and abnormal fluctuation correlation features;

[0008] Based on a preset anomaly detection strategy network, perform anomaly state matching processing on the state feature set to determine the anomaly level and anomaly influence range parameters of the operating equipment of the hydropower station;

[0009] Generate a set of equipment operation and maintenance optimization strategies according to the anomaly level and anomaly influence range parameters, and the set of equipment operation and maintenance optimization strategies includes an equipment control instruction sequence and a maintenance resource scheduling plan;

[0010] Feedback the set of equipment operation and maintenance optimization strategies to the hydropower station control center to trigger equipment operation and maintenance operations.

[0011] On the other hand, the present invention also provides an operation and maintenance management system for a hydropower station, including a processor and a machine-readable storage medium. 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.

[0012] Based on the above aspects, the present invention significantly improves the intelligent level and response efficiency of hydropower station operation and maintenance management by integrating real-time monitoring data of hydropower station operation equipment and constructing a dynamic feature extraction and anomaly detection strategy network. Specifically, by obtaining a set of real-time monitoring data containing sensor data streams of multiple monitoring time windows, and performing dynamic feature extraction operations on this basis, a set of state features covering multiple dimensions such as equipment operation stability, environmental coupling, and abnormal fluctuation correlation is generated, effectively mining the internal laws and potential risks of equipment operation. Further, based on a preset anomaly detection strategy network, in-depth matching processing is performed on the set of state features, which can accurately identify abnormal states and quantify their levels and influence ranges, breaking through the limitations of traditional operation and maintenance management that rely on manual experience or simple threshold judgment. Finally, according to the anomaly level and influence range parameters, a set of equipment operation and maintenance optimization strategies is dynamically generated and fed back to the hydropower station control center in real time to trigger operation and maintenance operations, not only realizing the optimal configuration and efficient scheduling of operation and maintenance resources, but also significantly enhancing the ability of the hydropower station to respond to sudden anomalies, thus contributing to the safe, stable and efficient operation of the hydropower station. Description of the Drawings

[0013] Figure 1 is a schematic execution flow diagram of an operation and maintenance management method for a hydropower station provided by an embodiment of the present invention.

[0014] Figure 2 is a schematic diagram of exemplary hardware and software components of an operation and maintenance management system for a hydropower station provided by an embodiment of the present invention. Detailed Embodiments

[0015] The present invention will be specifically described below in conjunction with the accompanying drawings of the specification. Figure 1 is a schematic flow diagram of an operation and maintenance management method for a hydropower station provided by an embodiment of the present invention. The operation and maintenance management method for a hydropower station will be introduced in detail below.

[0016] Step S110: Obtain a set of real-time monitoring data of hydropower station operation equipment, and the set of real-time monitoring data includes sensor data streams of multiple monitoring time windows.

[0017] In the hydropower station scenario of this embodiment, there are various operating devices in the hydropower station, such as power generation device X, water turbine device Y, and various auxiliary devices, etc. Around these operating devices, a plurality of sensors are arranged, such as vibration sensor A, temperature sensor B, pressure sensor C, and so on. These sensors continuously collect data to form their respective sensor data streams.

[0018] Set the duration of each monitoring time window to Δt. For example, starting from the starting time t0, the first monitoring time window is [t0, t0 + Δt], the second monitoring time window is [t0 + Δt, t0 + 2Δt], and so on. Within each monitoring time window, different types of sensors will generate corresponding data sequences. Taking vibration sensor A as an example, within the nth monitoring time window, the vibration amplitude data sequence V n =[V n1 ,V n2 ,…,V na is output, where a represents the number of vibration amplitude data points collected within this monitoring time window; temperature sensor B will output the temperature data sequence T n =[T n1 ,T n2 ,…,T nx , x is the number of temperature data points collected within this monitoring time window; pressure sensor C outputs the pressure data sequence P n =[P n1 ,P n2 ,…,P ne , e is the number of pressure data points collected within this monitoring time window. The real-time monitoring data set S contains all the data sequences generated by the sensors from the 1st monitoring time window to the Nth monitoring time window, that is, S = {(V1, T1, P1), (V2, T2, P2), …, (V n ,T n ,P n )}.

[0019] Step S120: Perform a dynamic feature extraction operation on the real-time monitoring data set to generate a state feature set of the hydropower station operating equipment, where the state feature set includes equipment operation stability features, environmental coupling features, and abnormal fluctuation correlation features.

[0020] To accurately monitor the state of the hydropower station operating equipment, it is necessary to perform dynamic feature extraction on the real-time monitoring data set to obtain a feature set that can effectively characterize the equipment state.

[0021] Step S121: For the sensor data streams of each monitoring time window, perform time-domain segmentation processing to obtain a plurality of data segment sets.

[0022] For the sensor data stream of each monitoring time window, time domain segmentation is performed according to the preset segment length τ (τ<Δt). Taking the vibration sensor data stream Vᵢ of the i-th monitoring time window as an example, it is divided into multiple data segments. Assuming that the number of data points in the monitoring time window is a, it can be divided into m=a / τ (if a is not divisible by τ, it is rounded down) data segments. The j-th data segment Vᵢⱼ=[Vᵢⱼ1, Vᵢⱼ2,…, Vᵢⱼ t ] (j=1, 2, …, m). Similarly, the temperature sensor data stream Tᵢ and the pressure sensor data stream Pᵢ are also subjected to the same time domain segmentation processing to obtain the temperature data segment set {Tᵢⱼ} and the pressure data segment set {Pᵢⱼ} respectively, thereby forming multiple data segment sets, such as the vibration sensor data segment set {Vᵢ1, Vᵢ2, …, Vᵢ m}, the set of temperature sensor data segments is {Tᵢ1, Tᵢ2, …, Tᵢ m}, the set of pressure sensor data segments is {Pᵢ1, Pᵢ2, …, Pᵢ m}.

[0023] Step S122: calling a preset feature coding model to perform multi-scale feature fusion processing on each data segment set to generate a time-frequency joint feature vector of the data segment set.

[0024] Step S1221: Perform fast Fourier transform processing on each data segment set to generate a frequency domain energy distribution matrix.

[0025] Taking the jth data segment Vᵢⱼ of the i-th monitoring time window of the vibration sensor as an example, a fast Fourier transform is performed on it. The time domain data is converted into frequency domain data through a predetermined algorithm to generate a frequency domain energy distribution matrix Fᵢⱼ. The elements Fᵢⱼ in the frequency domain energy distribution matrix k Represents the energy value at the kth frequency point. The number of rows and columns of the frequency domain energy distribution matrix depends on the length of the data segment and the frequency resolution set by the Fourier transform algorithm used. For example, assuming that the frequency resolution is set to be able to distinguish n frequency points from f1 to f2, then Fᵢⱼ is a matrix with n rows and 1 column, Fᵢⱼ=[Fᵢⱼ1, Fᵢⱼ2, …, Fᵢⱼ n ]. Similarly, the data segments of the temperature sensor and the pressure sensor are processed similarly to generate the temperature frequency domain energy distribution matrix Gᵢⱼ and the pressure frequency domain energy distribution matrix Hᵢⱼ, respectively.

[0026] Step S1222: extracting the energy peak sequence and energy mean parameter within a preset frequency band in the frequency domain energy distribution matrix.

[0027] For the frequency domain energy distribution matrix Fᵢⱼ of the vibration sensor, a frequency band range [f_low, f_high] is preset. In this frequency band, the peak value of the energy value is found to form an energy peak sequence E p For example, suppose there are n1 frequency points belonging to [f_low, f_high] within the frequency band, then E pi ⱼ=[E pi ⱼ1, E pi ⱼ2,…,E pi ⱼ n1 ], where E pi ⱼ k is the energy peak value corresponding to the kth frequency point in the frequency band. At the same time, the mean energy value within the frequency band is calculated to obtain the energy mean parameter E mi ⱼ. The way to calculate the mean here is to add up all the energy values in the frequency band and divide it by the number of frequency points, that is, E mi ⱼ = (E pi ⱼ1+E pi ⱼ2+…+E pi ⱼ n1 ) / n1. For the frequency domain energy distribution matrix Gᵢⱼ of the temperature sensor and the frequency domain energy distribution matrix Hᵢⱼ of the pressure sensor, the same operation is performed within their respective preset frequency bands to obtain the temperature energy peak sequence E pg ᵢⱼ、temperature energy mean parameter E mg ᵢⱼ and the pressure energy peak sequence E ph ᵢⱼ、Pressure energy mean parameter E mh ᵢⱼ.

[0028] Step S1223: Perform time domain difference calculation processing on the data segment set to generate a time domain change rate sequence and time domain mutation point position information.

[0029] Again, taking the jth data segment Vᵢⱼ of the i-th monitoring time window of the vibration sensor as an example, perform time domain difference calculation processing on it. Calculate the difference between adjacent data points to obtain the time domain change rate sequence Rᵢⱼ. For example, obtain the time interval Δt of the data segment (Δt=total duration of the monitoring time window / number of data points). Calculate the difference between adjacent data points to obtain the time domain change rate sequence Rᵢⱼ. For example, Rᵢⱼ1=(Vᵢⱼ2-Vᵢⱼ1) / Δt, Rᵢⱼ2=(Vᵢⱼ3-Vᵢⱼ2) / Δt, and so on, Rᵢⱼ=[Rᵢⱼ1, Rᵢⱼ2,…, Rᵢⱼ(t-1)]. At the same time, by setting a threshold θ1, when |Rᵢⱼ kWhen |>θ1, it is considered that there may be time-domain mutation points at this position. Record the position information of these mutation points to form the time-domain mutation point position information set Mᵢⱼ. The same processing is also performed on the data segments of the temperature sensor and the pressure sensor, respectively obtaining the temperature time-domain change rate sequence Rgᵢⱼ, the temperature time-domain mutation point position information set Mgᵢⱼ, the pressure time-domain change rate sequence Rhᵢⱼ, and the pressure time-domain mutation point position information set Mhᵢⱼ.

[0030] Step S1224: Perform standardization conversion on the energy peak sequence, the energy mean parameter, the time-domain change rate sequence, and the time-domain mutation point position information respectively, and then input them into the feature encoding model for feature cross-fusion processing to generate a fusion feature vector containing time-domain correlation weights and frequency-domain correlation weights.

[0031] For the energy peak sequence E related to the vibration sensor pij 、energy mean parameter E mi ⱼ, time-domain change rate sequence Rᵢⱼ, and time-domain mutation point position information Mᵢⱼ, perform standardization conversion. For the energy peak sequence E pi ⱼ, use the formula E pi ⱼ'=(E pi ⱼ-min(E pi ⱼ)) / (max(E pi ⱼ)-min(E pi ⱼ)) for standardization to obtain the standardized energy peak sequence E pi ⱼ'. For the energy mean parameter E mi ⱼ, perform standardization through the formula E mi ⱼ'=(E mi ⱼ-mean(E mi ⱼ)) / std(E mi ⱼ), where mean(E mi ⱼ) is the mean of E mi ⱼ, and std(E mi ⱼ) is the standard deviation of E mi ⱼ, to obtain the standardized energy mean parameter E mi ⱼ'. For the time-domain change rate sequence Rᵢⱼ, perform processing in a similar standardization manner as the energy peak sequence to obtain Rᵢⱼ'. For the time-domain mutation point position information Mᵢⱼ, convert it into a numerical form through a set encoding method, for example, map the position information to the interval [0, 1], and then perform standardization processing to obtain Mᵢⱼ'.

[0032] These standardized features are input into a pre-set feature encoding model. The feature encoding model performs cross-fusion processing on these features through its internal neural network structure or common algorithms of other related technologies. For example, different neuron layers in the model perform weighted processing on time-domain and frequency-domain features respectively, and finally generate a fused feature vector Fvᵢⱼ, which contains the time-domain correlation weight Wtᵢⱼ and the frequency-domain correlation weight Wfᵢⱼ. The same processing is also performed on the temperature sensor and the pressure sensor, respectively obtaining the fused feature vectors Fvgᵢⱼ and Fvhᵢⱼ, as well as the corresponding time-domain correlation weights Wtgᵢⱼ, Wtfᵢⱼ and Wthᵢⱼ, and frequency-domain correlation weights Wfgᵢⱼ, Wfhᵢⱼ and Wfhj.

[0033] Step S1225: Generate the time-frequency joint feature vector based on the similarity calculation between the fused feature vector and a pre-set reference feature template.

[0034] A reference feature template Bv is pre-set. For the fused feature vector Fvᵢⱼ generated by the vibration sensor, its similarity with the reference feature template Bv is calculated. For example, the cosine similarity algorithm can be used to calculate the cosine similarity cosine_simᵢⱼ=(Fvᵢⱼ·Bv) / (||Fvᵢⱼ||||Bv||), where "·" represents the vector dot product, and ||Fvᵢⱼ|| and ||Bv|| represent the vector norms of Fvᵢⱼ and Bv respectively. According to this similarity value, through the set mapping relationship, the time-frequency joint feature vector Tvᵢⱼ is generated. For example, if cosine_simᵢⱼ is greater than a certain threshold θ2, it is considered that the similarity between Fvᵢⱼ and Bv is high, and some elements of Tvᵢⱼ take the value of 1, otherwise take the value of 0, and the values of other elements of Tvᵢⱼ are determined according to other relevant rules. For the temperature sensor and the pressure sensor, the time-frequency joint feature vectors Tvgᵢⱼ and Tvhhᵢⱼ are also generated by calculating the similarity with the reference feature template Bv respectively.

[0035] Step S123: Construct the device operation stability feature based on the time-frequency joint feature vector. The device operation stability feature includes a vibration amplitude change trend parameter, a temperature fluctuation correlation parameter, and a pressure gradient matching parameter. The vibration amplitude change trend parameter is generated based on the peak sequence of the time-domain vibration signal and the frequency-domain energy integral value.

[0036] Taking a vibration sensor as an example, the time-frequency joint feature vector Tvᵢⱼ generated by it is used to construct the vibration amplitude change trend parameter. First, the peak sequence Psvᵢⱼ is extracted from the time-domain vibration signal. For example, by setting a peak detection threshold θ3, when the vibration amplitude is greater than this threshold and its adjacent amplitude is less than this amplitude, this point is considered a peak point, thus obtaining the peak sequence Psvᵢⱼ, whose unit is assumed to be mm / s². At the same time, the frequency-domain energy integral value Eivᵢⱼ is calculated, that is, the energy values of the frequency-domain energy distribution matrix Fᵢⱼ within the preset frequency band range [f_low, f_high] are integrated. Assuming that the frequency resolution within this frequency band range is Δf and the energy value is Fᵢⱼ k (k is the frequency point within this frequency band), originally Eivᵢⱼ = ∑(Fᵢⱼ k *Δf), and the unit is (mm / s²)²・Hz. To make the dimensions consistent, the frequency-domain energy integral is converted into an equivalent amplitude, that is, first take the square root of Eivᵢⱼ to get sqrt(Eivᵢⱼ), whose unit becomes mm / s²・sqrt(Hz), and then divide by the frequency resolution Δf to obtain the equivalent amplitude Eiv_ampᵢⱼ = sqrt(Eivᵢⱼ) / Δf, with the unit of mm / s². By performing a set processing on the peak sequence Psvᵢⱼ and the equivalent amplitude Eiv_ampᵢⱼ, here a calculation method based on the proportional relationship is adopted to generate the vibration amplitude change trend parameter Vtpᵢⱼ. For example, Vtpᵢⱼ = Psvᵢⱼ / Eiv_ampᵢⱼ (assuming Eiv_ampᵢⱼ is not 0, if it is 0, then other processing methods are adopted, such as setting a very small non-zero value to replace it). At this time, Vtpᵢⱼ is a dimensionless parameter and can reasonably characterize the equipment stability.

[0037] For a temperature sensor, based on the time-frequency joint feature vector Tvgᵢⱼ generated by it, combined with the historical change situation of the temperature data and the current time-frequency characteristics, the temperature fluctuation correlation parameter Tfpᵢⱼ is constructed. For example, by analyzing the temperature change situation on different time scales, the standard deviation Stgᵢⱼ of the temperature change is calculated, and then combined with some elements in the time-frequency joint feature vector, such as the time-domain correlation weight Wtgᵢⱼ and the frequency-domain correlation weight Wfgᵢⱼ, the temperature fluctuation correlation parameter is generated through the formula Tfpᵢⱼ = Stgᵢⱼ*Wtgᵢⱼ / Wfgᵢⱼ (assuming Wfgᵢⱼ is not 0, if it is 0, then other processing methods are adopted, such as setting a very small non-zero value to replace it).

[0038] For a pressure sensor, based on the time-frequency joint feature vector Tvhhᵢⱼ it generates, construct the pressure gradient matching parameter Pgpᵢⱼ. First, calculate the pressure gradient Gphᵢⱼ between adjacent monitoring points (or time points) of the pressure, with the unit assumed to be Pascal / second. Calculate the mean frequency-domain energy Emh_ampᵢⱼ, that is, average the energy values of the frequency-domain energy distribution matrix Hᵢⱼ within a preset frequency band range, with the unit being Pascal²·Hz. Redesign the formula, Pgpᵢⱼ = Gphᵢⱼ / Emh_ampᵢⱼ (assuming Emh_ampᵢⱼ is not 0, if it is 0, then adopt other processing methods, such as setting a very small non-zero value to replace it), which avoids unit conflicts and can reasonably reflect the actual relationship between pressure changes and energy. Combine the vibration amplitude change trend parameter Vtpᵢⱼ, the temperature fluctuation correlation parameter Tfpᵢⱼ, and the pressure gradient matching parameter Pgpᵢⱼ to form the equipment operation stability feature Sstᵢⱼ.

[0039] Step S124: Extract the difference information of the sensor data stream between adjacent monitoring time windows to generate an environmental coupling feature, and the environmental coupling feature includes a water flow velocity change influence factor, a water level dynamic correlation factor, and an external environmental disturbance coefficient.

[0040] Step S1241: Obtain the difference data distribution between the sensor data stream of the current monitoring time window and the sensor data stream of the previous monitoring time window.

[0041] Taking the vibration sensor as an example, assume the current monitoring time window is the i-th one, and the previous monitoring time window is the i - 1-th one. Obtain the difference data distribution between the vibration sensor data stream Vᵢ = [Vᵢ1, Vᵢ2,..., Vᵢ a of the i-th monitoring time window and the vibration sensor data stream Vᵢ₋1 = [Vᵢ₋ 11 , Vᵢ₋ 12 ,..., Vᵢ₋ 1a of the i - 1-th monitoring time window. Calculate the difference of each corresponding data point to obtain the difference data sequence Dv = [Dv1, Dv2,..., Dv a , where Dvk = Vᵢ k - Vᵢ₋ 1k . Similarly, for the temperature sensor and the pressure sensor, calculate the temperature difference data sequence Dt and the pressure difference data sequence Dp respectively.

[0042] Step S1242: Perform clustering analysis processing on the difference data distribution to determine the difference clustering center point and the clustering radius parameter.

[0043] Perform clustering analysis on the differential data sequence Dv of the vibration sensor. Adopt a clustering algorithm, such as the DBSCAN clustering algorithm. Analyze the differential data points through this algorithm and divide them into different clusters. Each cluster has a cluster center point Cv and a cluster radius Rv. For example, the calculation method of the cluster center point Cv is the mean vector of all data points in the cluster (the mean for one-dimensional data). For the differential data sequences Dt and Dp of the temperature sensor and the pressure sensor, perform the same clustering analysis, and respectively obtain the temperature cluster center point Ct, the cluster radius Rt, the pressure cluster center point Cp, and the cluster radius Rp.

[0044] Step S1243: After uniformly converting the unit of the differential cluster center point to the unit in the preset environmental impact factor mapping table, perform matching calculation to generate the water flow velocity change impact factor.

[0045] For the vibration sensor differential data cluster center point Cv, convert its unit to the same unit as that in the preset environmental impact factor mapping table. Assume that the unit related to vibration in the environmental impact factor mapping table is the set vibration energy density unit, while the current unit of Cv is the vibration amplitude unit, and perform conversion through the set conversion formula. After conversion, search for the item matching the converted value in the mapping table, and generate the water flow velocity change impact factor Wfv according to the corresponding relationship in the mapping table. For example, the mapping table stipulates that when the converted value is within a certain interval, the corresponding water flow velocity change impact factor is determined through a predefined functional relationship. Suppose there are multiple intervals in the mapping table, namely interval I1, interval I2, ……, interval In, and each interval corresponds to a functional relationship. When the converted cluster center point value Cv' falls within interval Ix, calculate the water flow velocity change impact factor Wfv through the function fx(Cv') corresponding to this interval. The function fx may be a function based on a linear relationship, a non-linear relationship, or other complex relationships. For example, the function fx may be a polynomial function, fx(Cv') = a1*Cv'^2 + a2*Cv' + a3, where a1, a2, and a3 are predefined coefficients, and these coefficients may be obtained through the analysis of historical data, experimental measurements, or theoretical derivations.

[0046] For the clustering center points Ct and Cp of the temperature sensor and the pressure sensor, unit conversion and matching calculations are also performed in the same way. The converted clustering center point value Ct' of the temperature sensor is matched with the mapping table. When Ct' falls within the interval Iy, the water flow velocity change influence factor Wft related to temperature is calculated through the function fy(Ct') of the corresponding interval. Similarly, for the pressure sensor, the converted clustering center point value Cp' is matched with the mapping table. When Cp' falls within the interval Iz, the water flow velocity change influence factor Wfp related to pressure is calculated through the function fz(Cp'). Finally, these three factors are combined by weighted splicing to form the final water flow velocity change influence factor Wf. Suppose the weights are w1, w2, and w3 respectively, and the weights here are preset according to the importance of each sensor's influence on the water flow velocity. First, normalize Wfv, Wft, and Wfp respectively to ensure that they have the same dimension and comparable scale. Let the normalized factors be Wfv', Wft', and Wfp' respectively. The normalization method can be to subtract their respective minimum values and divide by the range, that is, Wfv' = (Wfv - min(Wfv)) / (max(Wfv) - min(Wfv)), and the same applies to Wft' and Wfp'.

[0047] Then, the normalized factors are spliced according to the weights, Wf = [w1 * Wfv', w2 * Wft', w3 * Wfp'], and thus the water flow velocity change influence factor containing multi-dimensional information is obtained to comprehensively reflect the influence of different sensor data differences on the water flow velocity change.

[0048] Step S1244: Perform a correlation analysis based on the clustering radius parameter and the water level fluctuation range in the historical environmental data to generate a water level dynamic correlation factor.

[0049] For the clustering radius Rv of the vibration sensor, obtain the water level fluctuation range data within the corresponding time period from the historical environmental data. Assume that the water level fluctuation range recorded in the historical environmental data is a sequence L = [L1, L2, …, Ls] containing water level values at multiple time points, where Ls represents the water level value at time point ts. First, determine the historical time interval corresponding to the current monitoring time window. Suppose this interval is [ta, tb]. Find the subsequence of water level values La-b = [Lx, Lx+1, …, Ly] within the corresponding time interval in the historical water level fluctuation range data L, where Lx corresponds to time point ta and Ly corresponds to time point tb. Then, analyze the correlation between the clustering radius Rv and the water level subsequence La-b. A correlation analysis method can be adopted, such as calculating the Pearson correlation coefficient between them. Assume the vibration sensor clustering radius sequence is Rv_sequence = [Rv1, Rv2, …, Rvn] (if there are multiple clusters, consider multiple clustering radii), and the water level subsequence La-b = [Lx, Lx+1, …, Ly]. The calculation method of the Pearson correlation coefficient is as follows: First, calculate the means of Rv_sequence and La-b, denoted as mean_Rv and mean_L respectively. Then calculate their covariance cov(Rv_sequence, La-b) = sum((Rvi - mean_Rv) * (Lj - mean_L)) / (n - 1) (where i ranges from 1 to n and j ranges from x to y), as well as their respective standard deviations std_Rv and std_L. The Pearson correlation coefficient correlation_Rv-L = cov(Rv_sequence, La-b) / (std_Rv * std_L). Use this Pearson correlation coefficient to reflect the correlation between the vibration sensor clustering radius and the water level fluctuation. Similarly, for the clustering radius Rt of the temperature sensor and the clustering radius Rp of the pressure sensor, perform the same correlation analysis with the corresponding historical water level fluctuation range data, and obtain the correlation coefficient correlation_Rt-L between the temperature sensor and the water level fluctuation and the correlation coefficient correlation_Rp-L between the pressure sensor and the water level fluctuation respectively. Finally, combine these three correlation coefficients in a set way to generate the water level dynamic association factor Wl. For example, through weighted average, assume the weights are u1, u2, u3 respectively, Wl = u1 * correlation_Rv-L + u2 * correlation_Rt-L + u3 * correlation_Rp-L. Here, the weights u1, u2, u3 are also preset according to the importance of the correlation between each sensor and the water level fluctuation, so as to obtain a water level dynamic association factor that comprehensively reflects the correlation between the clustering radii of different sensors and the water level fluctuation.

[0050] Step S1245: Extract the timestamps and disturbance intensity parameters of the sudden disturbance events in the differential data distribution, and generate an external environment disturbance coefficient in combination with the external environment monitoring data.

[0051] For the differential data sequence Dv of the vibration sensor, a disturbance intensity threshold th4 is set to identify sudden disturbance events. When the absolute value of Dvk is greater than th4, it is considered that a sudden disturbance event occurs at the time point corresponding to this data point. Record the timestamps of these sudden disturbance events, denoted as the timestamp sequence Tstamps_v = [t1, t2,..., tm], where ti represents the time point when the i-th sudden disturbance event occurs. At the same time, take the absolute value of Dvk as the disturbance intensity parameter of this sudden disturbance event, denoted as the disturbance intensity sequence Strengths_v = [s1, s2,..., sm], where si represents the disturbance intensity of the i-th sudden disturbance event. Similarly, for the differential data sequences Dt and Dp of the temperature sensor and the pressure sensor, the same sudden disturbance event identification is performed, and the timestamp sequence Tstamps_t = [t1', t2',..., tn] and the disturbance intensity sequence Strengths_t = [s1', s2',..., sn] of the temperature sensor, and the timestamp sequence Tstamps_p = [t1'', t2'',..., to] and the disturbance intensity sequence Strengths_p = [s1'', s2'',..., so] of the pressure sensor are obtained respectively.

[0052] The external environment monitoring data may come from various external environment monitoring devices arranged around the hydropower station, such as wind speed and wind direction data monitored by a weather station, seismic wave data monitored by seismic monitoring equipment, etc. Let the wind speed data sequence be Wind_speed = [ws1, ws2,..., wp], the wind direction data sequence be Wind_direction = [wd1, wd2,..., wp], the seismic wave intensity data sequence be Seismic_intensity = [si1, si2,..., sq], etc.

[0053] Combine these external environmental monitoring data and the information of sudden disturbance events of sensors to generate the external environmental disturbance coefficient We. For the vibration sensor, taking the time - stamp sequence Tstamps_v as an example, find the corresponding external environmental data for each time - stamp ti in the external environmental monitoring data. For example, find the wind - speed value wsi' corresponding to the time - point ti in the wind - speed data sequence Wind_speed, find the corresponding wind - direction value wdi' in the wind - direction data sequence Wind_direction, and find the corresponding seismic - wave intensity value sii' in the seismic - wave intensity data sequence Seismic_intensity. Calculate the external environmental disturbance contribution value We_v corresponding to the vibration sensor through a set function relationship, such as a multivariate function fe(wsi', wdi', sii', si). The function fe may comprehensively consider the influence of wind speed, wind direction, seismic - wave intensity, and disturbance intensity on the external environmental disturbance, and its specific form may be obtained through learning from historical data or theoretical analysis. Similarly, for the temperature sensor and the pressure sensor, calculate the external environmental disturbance contribution values We_t and We_p corresponding to the temperature sensor and the pressure sensor respectively. Finally, combine these three contribution values through weighted splicing to generate the external environmental disturbance coefficient We. Assuming the weights are v1, v2, and v3 respectively, first normalize We_v, We_t, and We_p. Let the normalized factors be We_v', We_t', and We_p' respectively, and the normalization method is similar to the previous one. Then splice the normalized factors according to the weights, We = [v1*We_v', v2*We_t', v3*We_p'], so as to obtain an external environmental disturbance coefficient that comprehensively reflects the relationship between sudden disturbance events of different sensors and external environmental data.

[0054] Step S125: After standardizing the device operation stability feature, the environmental coupling feature, and the abnormal fluctuation correlation feature respectively, merge them into the state - feature set.

[0055] For the device operation stability feature Sstij, it includes the vibration amplitude change trend parameter Vtpij, the temperature fluctuation correlation parameter Tfpij, and the pressure gradient matching parameter Pgpij. Standardize these three parameters respectively. Among them, before the standardization process, the vibration parameter Vtpij is converted to a decibel value: Vtp_dB = 20 * log10(Vtpij / V_ref), where V_ref is the reference amplitude. The temperature parameter Tfpij is converted to a normalized temperature difference: Tfp_norm = (Tfpij - T_min) / (T_max - T_min), and T_min and T_max are the historical temperature extreme values. The pressure parameter Pgpij is converted to a dimensionless ratio: Pgp_ratio = Pgpij / P_baseline, where P_baseline is the reference pressure value. For Vtpij, perform Z-score standardization to obtain the standardized vibration amplitude change trend parameter Vtpij'. Similar standardization processes are performed on Tfpij and Pgpij to obtain Tfpij' and Pgpij' respectively. Combine these three standardized parameters to form the standardized device operation stability feature Sstij' = [Vtpij', Tfpij', Pgpij'].

[0056] For the environmental coupling feature, it includes the water flow velocity change influence factor Wf, the water level dynamic correlation factor Wl, and the external environmental disturbance coefficient We. Standardize each dimension in Wf. Let Wf = [w1 * Wfv', w2 * Wft', w3 * Wfp'], and standardize w1 * Wfv', w2 * Wft', and w3 * Wfp' respectively. For example, for w1 * Wfv', use the formula (w1 * Wfv')' = ((w1 * Wfv') - min(w1 * Wfv')) / (max(w1 * Wfv') - min(w1 * Wfv')) for standardization to obtain the standardized (w1 * Wfv')'. Similarly, standardize w2 * Wft' and w3 * Wfp' to obtain the standardized (w2 * Wft')' and (w3 * Wfp')'. Recombine these three standardized dimensions to form the standardized water flow velocity change influence factor Wf' = [(w1 * Wfv')', (w2 * Wft')', (w3 * Wfp')']. Similar standardization processes are performed on the water level dynamic correlation factor Wl and the external environmental disturbance coefficient We to obtain Wl' and We' respectively. Combine these three standardized parts to form the standardized environmental coupling feature Ec' = [Wf', Wl', We'].

[0057] For the abnormal fluctuation correlation features (the extraction process of the abnormal fluctuation correlation features is not described in detail here. Assume that it has been extracted in a similar detailed manner as before and is denoted as Abij, which includes the sudden fluctuation frequency Fbij, the fluctuation duration correlation parameter Dbij, and the fluctuation propagation path topological relationship Tbij), standardize the sudden fluctuation frequency Fbij. Use the formula Fbij'=(Fbij - min(Fbij)) / (max(Fbij) - min(Fbij)) to obtain the standardized sudden fluctuation frequency Fbij'. Standardize the fluctuation duration correlation parameter Dbij. Assume that Dbij is a parameter set containing multiple dimensions of information, and process each dimension in a similar standardization manner to obtain the standardized fluctuation duration correlation parameter Dbij'. For the fluctuation propagation path topological relationship Tbij, convert it into a numerical form through a set encoding method, and then perform standardization processing to obtain the standardized fluctuation propagation path topological relationship Tbij'. Combine these three standardized parts to form the standardized abnormal fluctuation correlation feature Abij' = [Fbij', Dbij', Tbij'].

[0058] Finally, merge the standardized device operation stability feature Sstij', the environmental coupling feature Ec', and the abnormal fluctuation correlation feature Abij' into the state feature set St = [Sstij', Ec', Abij']. In this way, the conversion from the real-time monitoring data set to the state feature set is completed, providing a basis for subsequent anomaly detection and operation and maintenance strategy formulation.

[0059] Step S130: Based on a preset abnormal detection strategy network, perform abnormal state matching processing on the state feature set to determine the abnormal level and abnormal impact range parameters of the hydropower station operation equipment.

[0060] The preset abnormal detection strategy network is a trained network model that can analyze the input state feature set to determine information related to the abnormal state of the equipment.

[0061] Step S131: Input the state feature set into the abnormal detection strategy network for feature importance ranking processing to generate a feature weight distribution vector.

[0062] Input the set of state features St = [Sstij', Ec', Abij'] into the anomaly detection policy network. Some algorithms may be adopted inside the anomaly detection policy network to evaluate the importance of each feature in judging the abnormal state of the device. For example, it may learn from a large amount of historical data (including device data and corresponding state feature sets under normal and abnormal states) that certain features are more critical for judging anomalies. Taking the vibration amplitude change trend parameter Vtpij' in the device operation stability feature Sstij' as an example, the network may analyze the correlation between the change pattern of Vtpij' and the occurrence of anomalies in the historical data. By performing similar analyses on all features, the features are sorted according to their importance. Assume that all features included in the set of state features St are arranged in order as f1, f2,..., fn. The network assigns a weight to each feature according to the importance analysis, forming a feature weight distribution vector W = [w1, w2,..., wn], where wi represents the weight of feature fi. The magnitude of the weight reflects the relative importance of the feature in judging the abnormal state of the device. The larger the weight, the greater the impact of the feature on the anomaly judgment.

[0063] Step S132: Perform weighted fusion processing on the set of state features based on the feature weight distribution vector to generate a comprehensive device state score.

[0064] Use the generated feature weight distribution vector W = [w1, w2,..., wn] to perform weighted fusion processing on the set of state features St. For each feature fi in the set of state features St, multiply it by the corresponding weight wi, and then sum all the products to obtain the comprehensive device state score Score. That is, Score = w1*f1 + w2*f2 +... + wn*fn. This weighted fusion process takes into account the importance of each feature, enabling the comprehensive score to more accurately reflect the overall state of the device. For example, if the weight w1 of the vibration amplitude change trend parameter Vtpij' is large, it indicates that the vibration amplitude change trend parameter has a more significant impact when judging the abnormal state of the device, and thus it accounts for a larger proportion in the comprehensive score.

[0065] Step S133: Dynamically generate a dynamic anomaly threshold interval adaptively adjusted based on the device operation state according to the type of the hydropower station operation device and the historical normal operation data. Compare the comprehensive device state score with the dynamic anomaly threshold interval. If the comprehensive device state score exceeds the dynamic anomaly threshold interval, trigger the anomaly level classification process.

[0066] Different types of hydropower station operating equipment have different normal operating parameter ranges and characteristics. First, determine the type of equipment being processed currently, assumed to be equipment type X. Obtain the state characteristic data of equipment type X in the normal operating state from historical normal operating data. Assume that the historical normal operating data contains state characteristic sets at multiple time points, namely St1, St2, …, Stm. Analyze these historical state characteristic sets to determine the dynamic anomaly threshold interval. For example, the mean mean_fi and standard deviation std_fi of each characteristic fi in the historical normal operating data can be calculated. Based on these statistical information, determine the dynamic anomaly threshold interval through a predefined functional relationship. Assume that the lower limit of the dynamic anomaly threshold interval is Lower_bound = mean_fi - k1 * std_fi, and the upper limit is Upper_bound = mean_fi + k2 * std_fi, where k1 and k2 are predefined coefficients, and these coefficients may be determined according to factors such as the characteristics of the equipment, the fluctuation of historical data, and the sensitivity requirements for anomaly judgment. For each characteristic fi in the state characteristic set St, determine its corresponding dynamic anomaly threshold interval in this way. Then, compare the comprehensive equipment state score Score with these dynamic anomaly threshold intervals. If Score exceeds all the dynamic anomaly threshold intervals corresponding to the characteristics (i.e., Score is less than the Lower_bound of all characteristics or greater than the Upper_bound of all characteristics), it is considered that the equipment state is abnormal, and the abnormal level classification process is triggered.

[0067] For another example, based on the historical data distribution, the mean μ and standard deviation σ of each characteristic fi can also be calculated. According to the equipment type and risk level, select the dynamic coefficient α (default α = 3):

[0068] Lower_bound = μ - ασ,

[0069] Upper_bound = μ + ασ.

[0070] If the comprehensive equipment state score Score exceeds [Lower_bound, Upper_bound], the abnormal level classification is triggered.

[0071] Step S134: Call the preset abnormal impact assessment model, and calculate the abnormal impact range parameter according to the abnormal fluctuation correlation feature and environmental coupling feature in the state characteristic set. The abnormal impact range parameter includes the equipment component correlation parameter, the upstream and downstream equipment impact coefficient, and the maintenance priority score.

[0072] The preset abnormal impact assessment model is a model specifically used to evaluate the impact scope of abnormalities on equipment. Taking the topological relationship Tbij' of the fluctuation propagation path in the abnormal fluctuation correlation feature Abij' as an example, the model first analyzes the connection relationship between the equipment components represented by Tbij'. Suppose Tbij' can be represented as a graph structure, where the nodes represent the equipment components and the edges represent the connection relationships between the components. By analyzing this graph structure, a graph theory-based algorithm is adopted, such as calculating indicators like the shortest path length between nodes and the degree of nodes, to determine the degree of connection tightness between the equipment components. Let the set of equipment components be C = [c1, c2,..., cl]. By analyzing the graph structure, an association degree value Rij is calculated for each pair of components (ci, cj). Then, based on these association degree values, the equipment component association degree parameter is generated. For example, by performing a certain statistical analysis on all the association degree values Rij, such as calculating the mean, median, or using a weighted average method, the equipment component association degree parameter Ccp can be obtained.

[0073] For the upstream and downstream equipment impact coefficients, the water flow velocity change impact factor Wf' and the water level dynamic correlation factor Wl' in the environmental coupling feature Ec' are combined. Suppose there is a hydraulic connection between the hydropower station equipment, and the changes in water flow velocity and water level will affect the upstream and downstream equipment. By analyzing the relationship between the water flow velocity, water level, and the operating states of the upstream and downstream equipment in historical data, a functional relationship fu(Wf', Wl') is established to calculate the upstream and downstream equipment impact coefficient Ue. The function fu may consider factors such as the impact of water flow velocity change on the flow rate of the upstream and downstream equipment, and the impact of water level change on the pressure of the upstream and downstream equipment. For example, an increase in water flow velocity may lead to an increase in the flow rate of the downstream equipment. By fitting and analyzing the historical data, the quantitative relationship between the water flow velocity change and the downstream equipment flow rate change is determined, and this relationship is reflected in the function fu.

[0074] For the maintenance priority score, the equipment component association degree parameter Ccp and the upstream and downstream equipment impact coefficient Ue are comprehensively considered. A weighted evaluation algorithm is used to generate the maintenance priority score Mps. Let the weights be q1 and q2 respectively, Mps = q1 * Ccp + q2 * Ue. Here, the weights q1 and q2 are preset according to the importance of the equipment component association degree and the upstream and downstream equipment impact on the maintenance priority. Finally, the equipment component association degree parameter Ccp, the upstream and downstream equipment impact coefficient Ue, and the maintenance priority score Mps are combined to form the abnormal impact scope parameter Ar = [Ccp, Ue, Mps]. It should be noted here that since the equipment component association degree parameter Ccp, the upstream and downstream equipment impact coefficient Ue, and the maintenance priority score Mps may have different dimensions or value ranges, for the consistency and rationality of subsequent processing, they may need to be further processed.

[0075] For the equipment component correlation parameter Ccp, if it is obtained through a certain statistical analysis of the correlation value Rij and its value range is quite different from the upstream and downstream equipment influence coefficient Ue and the maintenance priority score Mps, normalization can be adopted. Assume the value range of Ccp is [min_Ccp, max_Ccp], and it is normalized through the formula Ccp'=(Ccp - min_Ccp) / (max_Ccp - min_Ccp) to obtain the normalized equipment component correlation parameter Ccp'.

[0076] For the upstream and downstream equipment influence coefficient Ue, if it does not match the other two parameters in terms of dimension or value range, corresponding processing is also carried out. For example, if Ue is the comprehensive calculation result of the influence of water flow velocity and water level change on upstream and downstream equipment and its value range is in a specific interval [min_Ue, max_Ue], it is normalized through the formula Ue'=(Ue - min_Ue) / (max_Ue - min_Ue) to obtain the normalized upstream and downstream equipment influence coefficient Ue'.

[0077] For the maintenance priority score Mps, if there are differences in dimension or value range from the previous two, similar normalization operations are also taken. Assume its value range is [min_Mps, max_Mps], and after normalization through the formula Mps'=(Mps - min_Mps) / (max_Mps - min_Mps), the normalized maintenance priority score Mps' is obtained.

[0078] The three normalized parameters are recombined to form the final abnormal influence range parameter Ar' = [Ccp', Ue', Mps']. After such processing, each part of the abnormal influence range parameter has better consistency in dimension and value range, which is convenient for subsequent analysis and application.

[0079] Step S135: Determine the abnormal level according to the deviation degree of the comprehensive equipment state score from the abnormal threshold interval and the abnormal influence range parameter.

[0080] First, analyze the deviation degree of the comprehensive equipment state score Score from the dynamic abnormal threshold interval [Lower_bound, Upper_bound]. Calculate the differences between Score and Lower_bound and Upper_bound, which are Deviation_lower = Score - Lower_bound (when Score < Lower_bound) or Deviation_upper = Score - Upper_bound (when Score > Upper_bound).

[0081] The abnormal level is determined based on the deviation degree and the maintenance priority score Mps' in the abnormal influence range parameter Ar'. Assume that multiple abnormal level intervals are preset in advance, such as level 1, level 2, level 3, etc. Each level interval corresponds to different deviation degree ranges and maintenance priority score ranges.

[0082] Taking a simple division method as an example, when the absolute value of Deviation_lower is less than a certain threshold th5 and Mps' is less than a certain threshold th6, the abnormal level is determined to be level 1, indicating a lower degree of abnormality. When the absolute value of Deviation_lower is greater than or equal to th5 and less than th7, and at the same time Mps' is greater than or equal to th6 and less than th8, the abnormal level is determined to be level 2, representing a medium degree of abnormality. When the absolute value of Deviation_upper is greater than or equal to th7 and Mps' is greater than or equal to th8, the abnormal level is determined to be level 3, meaning a higher degree of abnormality. The thresholds th5, th6, th7, th8, etc. are preset based on various factors such as the historical operation data of the hydropower station equipment, the degree of fault influence, and the operation and maintenance experience, so as to ensure that the division of the abnormal level can accurately reflect the actual situation of the equipment abnormality. In this way, by comprehensively considering the deviation degree of the comprehensive equipment status score from the abnormal threshold interval and the abnormal influence range parameter, an accurate abnormal level is determined, providing an important basis for formulating corresponding operation and maintenance optimization strategies in the future.

[0083] Step S140: Generate a set of equipment operation and maintenance optimization strategies according to the abnormal level and the abnormal influence range parameter. The set of equipment operation and maintenance optimization strategies includes an equipment control instruction sequence and a maintenance resource scheduling plan.

[0084] After determining the abnormal level and the abnormal influence range parameter of the equipment, the next step is to generate a corresponding set of equipment operation and maintenance optimization strategies to guide the maintenance and management of the hydropower station operation equipment.

[0085] Step S141: Match the preset operation and maintenance response strategy template according to the abnormal level to generate a basic control instruction sequence.

[0086] Operation and maintenance response strategy templates corresponding to different abnormal levels are preset in advance. Assume that the abnormal levels are divided into level 1, level 2, and level 3, and each level has a corresponding template. Taking the abnormal level of level 2 as an example, the corresponding operation and maintenance response strategy template stipulates a series of operation guides for this level of abnormality.

[0087] The control measures to be taken for possible abnormal conditions of different device components are specified in detail in these templates. For example, for power generation device X, if an abnormal condition of level 2 occurs, the template stipulates that its output power needs to be adjusted. Specifically, the template may specify that the power adjustment parameter P of power generation device X needs to be adjusted to a certain range [P1, P2]. The adjustment of this power adjustment parameter P may involve a series of control instructions, such as changing the excitation current of the generator and adjusting the guide vane opening of the water turbine. Arrange these specific control instructions in the order of execution to form a basic control instruction sequence. Assume that the basic control instruction sequence is [instruction 1, instruction 2,..., instruction n], where instruction 1 may be "adjust the excitation current of the generator to value I1", instruction 2 may be "adjust the guide vane opening of the water turbine to angle A1", etc. Each instruction clearly stipulates the operation object, the content of the operation, and the parameter value to be achieved, thus constituting the basic control instruction sequence for the abnormal condition of level 2 and providing a basic framework for subsequent further optimization.

[0088] Step S142: Based on the maintenance priority score in the abnormal influence range parameter, perform dynamic adjustment processing on the basic control instruction sequence to generate an optimized device control instruction sequence.

[0089] The maintenance priority score Mps' in the abnormal influence range parameter reflects the priority degree of different device components or device-related parts in maintenance. Taking the device component with a higher maintenance priority score Mps' as an example, assume that a certain key component C1 in power generation device X has a higher maintenance priority because it has an important impact on the stability and safety of the entire power generation process.

[0090] In the basic control instruction sequence, for the control instructions related to this component C1, perform dynamic adjustment according to the maintenance priority score. For example, the original instruction for component C1 in the basic control instruction sequence is "perform a routine inspection on component C1 at time t1". Due to its higher maintenance priority, according to the maintenance priority score Mps' and the preset time adjustment rule, the time t1 may be advanced to t1', that is, adjusted to "perform a routine inspection on component C1 at time t1'", where t1' < t1. At the same time, the relevant operation parameters of component C1 may also be adjusted. Assume that the original instruction sets a certain operating parameter of component C1 to value V1. Considering its high maintenance priority and the current device state, according to the maintenance priority score Mps' and the parameter adjustment rule, the parameter value is adjusted to V1' to make the operating state of component C1 more in line with the maintenance requirements under the current abnormal condition.

[0091] For the upstream and downstream equipment impact coefficient Ue, if it indicates that a certain downstream equipment Y is greatly affected by the current anomaly, it may be necessary to insert a collaborative control instruction for the downstream equipment Y in the basic control instruction sequence. For example, insert the instruction "Adjust the opening of the flow control valve of the downstream equipment Y to B1 to maintain the overall operation stability of the system". By operating on the execution time, parameters of the instructions in the basic control instruction sequence and inserting collaborative control instructions, etc., and comprehensively considering the maintenance priority score and the upstream and downstream equipment impact coefficient, an optimized equipment control instruction sequence is generated to make it more adaptable to the current abnormal situation of the equipment and improve the efficiency and effect of operation and maintenance.

[0092] Step S143: Extract the equipment component correlation parameter and the upstream and downstream equipment impact coefficient from the abnormal influence range parameter, and generate a maintenance resource scheduling plan in combination with the maintenance resource inventory data. The maintenance resource scheduling plan includes a spare part requirement list, a personnel scheduling path, and a maintenance time window allocation parameter.

[0093] Extract the equipment component correlation parameter Ccp' and the upstream and downstream equipment impact coefficient Ue' from the abnormal influence range parameter Ar'. The equipment component correlation parameter Ccp' reflects the degree of tight correlation between each equipment component. Based on this, a set of target components that need to be preferentially maintained is selected. For example, by setting a correlation threshold th9, when the correlation value Rij (Ccp' is calculated based on these Rij) between components is greater than th9, the corresponding components are included in the set of target components. Assume the set of target components is {T1, T2,..., Tm}, where T1, T2, etc. represent different equipment components.

[0094] For each component in the target component set, a spare parts requirement list is dynamically generated based on its historical fault data, current state parameters, and the fluctuation propagation path topological relationship Tbij' in the abnormal fluctuation association feature Abij'. Taking component T1 as an example, its historical fault data is analyzed to find that in similar abnormal situations, the sub-components that often fail are S1, S2, etc. According to the abnormal conditions of the current state parameters of component T1, such as temperature, pressure, vibration and other parameters, the type and quantity of spare parts that may need to be replaced are further determined. Assume that according to the analysis, it is determined that the required spare parts types are Sp1 and Sp2, and the quantities are N1 and N2 respectively. At the same time, combined with the fluctuation propagation path topological relationship Tbij', if it is found that the fault may propagate to other components along a specific path, it may be necessary to prepare some additional related spare parts, such as spare part type Sp3 and quantity N3. After determining the type and quantity of spare parts, the emergency allocation path is determined according to the storage location and urgency of the spare parts. For example, spare part Sp1 is stored in warehouse W1. Due to its high urgency, it is determined to be urgently allocated from warehouse W1 through transportation route R1. The spare parts demand information of all target components is aggregated to form a spare parts demand list. The list format may be [(Sp1, N1, R1), (Sp2, N2, R2), (Sp3, N3, R3), …], where (Sp1, N1, R1) represents the spare parts type Sp1, quantity N1, and emergency allocation path R1.

[0095] Determine the list of affected external devices based on the upstream and downstream device influence coefficient Ue'. Assume that Ue' indicates that downstream devices Y1, Y2 and upstream device Z1 are affected by the current abnormality. Combined with the maintenance personnel location data, generate a personnel scheduling path. The maintenance personnel location data records the current location of each maintenance personnel, such as maintenance personnel P1 at location L1, maintenance personnel P2 at location L2, etc. Plan the personnel scheduling path based on the location of the affected equipment and the location of the maintenance personnel, taking into account the path priority and time estimation parameters. For example, for downstream device Y1, since its failure may have a greater impact on the operation of the entire system, the path priority is set to high. Maintenance personnel P1 is close to Y1 and has the skills to handle Y1 failures, so the planned personnel scheduling path is for P1 to start from location L1 and go to device Y1 via route R4, with an estimated arrival time of t2. For other affected equipment, the maintenance personnel and scheduling paths are determined in a similar manner to form a personnel scheduling path list, such as [(P1, Y1, R4, t2), (P3, Y2, R5, t3), (P4, Z1, R6, t4)], where (P1, Y1, R4, t2) means that maintenance personnel P1 goes to equipment Y1 via route R4, with an estimated arrival time of t2.

[0096] Based on the maintenance time window allocation parameters, perform spatio-temporal conflict detection and processing on the spare part allocation path and the personnel scheduling path. The maintenance time window allocation parameters specify the time range during which each maintenance task can be executed. For example, for the maintenance task of component T1, the maintenance time window is [tw1, tw2]. Check whether there are conflicts between the spare part allocation path and the personnel scheduling path in terms of time and space. Suppose there is a spatial conflict (such as the two paths crossing in the same area) between the allocation path R1 of spare part Sp1 and the path R4 of maintenance personnel P1 going to equipment Y1 during a certain time period [ts1, ts2], and this time period is within the maintenance time window [tw1, tw2] of component T1. According to the maintenance priority score Mps', if the maintenance priority of component T1 is higher than that of equipment Y1, dynamically adjust the personnel scheduling path. For example, adjust the path of maintenance personnel P1 going to equipment Y1 to R7 to avoid spatio-temporal conflicts. If there are conflicts between the spare part allocation path and multiple personnel scheduling paths, comprehensively consider adjusting the spare part allocation path or the personnel scheduling path according to the maintenance priority score and other preset rules to generate maintenance time window allocation parameters without spatio-temporal conflicts. For example, after adjustment, the maintenance time window allocation parameters may become [(T1, [tw1, tw2], R1, P5), (Y1, [tw3, tw4], R7, P1),...], where (T1, [tw1, tw2], R1, P5) means that component T1 is within the maintenance time window [tw1, tw2], obtains spare parts through the spare part allocation path R1, and is maintained by maintenance personnel P5. Integrate the spare part requirement list, the personnel scheduling path, and the maintenance time window allocation parameters into a maintenance resource scheduling plan.

[0097] Step S144: Align the device control instruction sequence with the maintenance resource scheduling plan on the time axis to maintain the timing consistency between the execution of the device control instruction and the resource scheduling operation.

[0098] Both the device control instruction sequence and the maintenance resource scheduling plan involve time-related information. Each instruction in the device control instruction sequence has its expected execution time. For example, the instruction "adjust the excitation current of the generator to value I1 at time t1". The personnel scheduling path and the spare part allocation path in the maintenance resource scheduling plan also have corresponding time arrangements. For example, maintenance personnel P1 is expected to arrive at equipment Y1 at time t2 for maintenance, and spare part Sp1 is expected to reach the target component T1 through the allocation path R1 at time t3.

[0099] To maintain the timing consistency between the execution of device control instructions and resource scheduling operations, a unified timeline is first established. Taking the execution time of the first instruction in the device control instruction sequence as the starting reference point, assumed to be time t0. For each operation in the maintenance resource scheduling plan, it is adjusted according to its estimated time relationship with time t0. For example, if the estimated arrival time t2 of maintenance personnel P1 at device Y1 is earlier than the execution time t4 of a certain instruction in the device control instruction sequence for device Y1, and the execution of this instruction depends on the arrival of the maintenance personnel and certain preparatory work, then it may be necessary to adjust the scheduling path of the maintenance personnel or the execution time of the device control instruction. If it is not feasible to adjust the scheduling path of the maintenance personnel, the execution time t4 of the device control instruction can be postponed to time t4' after the maintenance personnel P1 arrives and completes the preparatory work.

[0100] For the spare part transfer path, if the estimated arrival time t3 of spare part Sp1 at the target component T1 is later than the execution time t5 of a certain instruction in the device control instruction sequence for component T1 that requires replacement of the spare part, and this instruction cannot be executed without the spare part, then it is necessary to speed up the transfer speed of the spare part or adjust the execution time of the device control instruction. For example, the transfer of spare part Sp1 can be accelerated by selecting a faster transportation method so that it can arrive before time t5, or the execution time t5 of the device control instruction can be postponed to time t5' after the spare part arrives. By carefully comparing and adjusting the time-related information in the device control instruction sequence and the maintenance resource scheduling plan in this way, it is ensured that during the entire operation and maintenance process, the execution of device control instructions and resource scheduling operations are coordinated with each other in terms of time, avoiding situations such as instructions not being executable due to unarrived resources or resource waste due to premature execution of instructions, thereby improving the efficiency and effectiveness of operation and maintenance operations.

[0101] Step S145: Combine the aligned device control instruction sequence and the maintenance resource scheduling plan into the device operation and maintenance optimization strategy set.

[0102] After completing the timeline alignment process of the device control instruction sequence and the maintenance resource scheduling plan, the two are combined into a device operation and maintenance optimization strategy set. The aligned device control instruction sequence is arranged in chronological order and contains specific control operations and execution times for abnormal situations of hydropower station operation equipment. For example, the device control instruction sequence may be [execute instruction 1 (adjust the generator excitation current to value I1) at time t1, execute instruction 2 (adjust the turbine guide vane opening to angle A1) at time t2,...].

[0103] The maintenance resource scheduling plan details the spare parts, personnel required for maintenance, as well as the corresponding scheduling paths and time arrangements, such as [spare parts demand list: (Sp1, N1, R1), (Sp2, N2, R2); personnel scheduling path: (P1, Y1, R4, t2), (P3, Y2, R5, t3); maintenance time window allocation parameters: (T1, [tw1, tw2], R1, P5), (Y1, [tw3, tw4], R7, P1)].

[0104] Combining these two parts together forms a set of equipment operation and maintenance optimization strategies. This set of equipment operation and maintenance optimization strategies comprehensively covers a series of optimization strategies from equipment control operations to maintenance resource allocation when equipment anomalies occur, providing detailed, orderly, and operable guidance for the operation and maintenance personnel of the hydropower station to ensure that equipment anomalies can be handled promptly and effectively, and to guarantee the stable operation of the hydropower station. For example, the set of equipment operation and maintenance optimization strategies can be expressed as {equipment control instruction sequence: [execute instruction 1 (adjust the generator excitation current to value I1) at time t1, execute instruction 2 (adjust the water turbine guide vane opening to angle A1) at time t2,...]; maintenance resource scheduling plan: [spare parts demand list: (Sp1, N1, R1), (Sp2, N2, R2); personnel scheduling path: (P1, Y1, R4, t2), (P3, Y2, R5, t3); maintenance time window allocation parameters: (T1, [tw1, tw2], R1, P5), (Y1, [tw3, tw4], R7, P1)]}.

[0105] Step S150: Feed back the set of equipment operation and maintenance optimization strategies to the hydropower station control center to trigger equipment operation and maintenance operations.

[0106] After the set of equipment operation and maintenance optimization strategies is generated, it needs to be fed back to the hydropower station control center to trigger actual equipment operation and maintenance operations.

[0107] For example, step S151: Parse the equipment control instruction sequence in the set of equipment operation and maintenance optimization strategies to generate executable control signal codes.

[0108] The device control instruction sequence contains a series of operation instructions for the operating devices of the hydropower station, such as "Adjust the generator excitation current to value I1", "Adjust the opening of the turbine guide vane to angle A1", etc. Analyze these instructions and convert them into executable control signal codes. Taking the instruction "Adjust the generator excitation current to value I1" as an example, first determine the device interface and communication protocol corresponding to this instruction. Assume that the control interface of the generator follows a certain set communication protocol, such as the Modbus protocol. According to the regulations of this protocol, convert the parameters (such as value I1) and operation type (adjust the excitation current) in the instruction into the corresponding Modbus register address and function code. For example, by querying the Modbus register mapping table, determine that the register address corresponding to adjusting the excitation current is Reg1 and the function code is Func1. Encode value I1 according to the data format of the Modbus protocol, such as converting it into the hexadecimal data format. Finally, according to the frame structure of the Modbus protocol, construct a control signal code frame containing the register address, function code, and encoded parameters. Assume that the format of the control signal code frame is [start bit, device address, function code, register address, data length, data content, check bit, end bit], and fill the above determined information into the control signal code frame to obtain the complete executable control signal code. For the instruction "Adjust the opening of the turbine guide vane to angle A1", follow a similar process. First, determine the communication protocol followed by the turbine control interface, assume it is another protocol Protocol2. By querying the register mapping table corresponding to this protocol, determine that the register address corresponding to adjusting the guide vane opening is Reg2 and the function code is Func2. Encode angle A1 according to the data format of Protocol2 protocol, and then construct a control signal code frame containing the register address, function code, and encoded parameters according to the frame structure of Protocol2 protocol. For each instruction in the device control instruction sequence, perform such a detailed analysis and code generation process to ensure that each device operation instruction can be converted into a control signal code recognizable and executable by the corresponding device. These generated control signal codes will be used to directly control the operating devices of the hydropower station in the follow-up, so that they can be adjusted and operated according to the predetermined operation and maintenance optimization strategy.

[0109] Step S152: Inject the control signal code into the device control interface of the hydropower station control center to trigger the operation parameter adjustment operation of the device.

[0110] The hydropower station control center is equipped with various device control interfaces, which are the bridges connecting the control signal codes and the actual operating devices. Taking the power generation equipment as an example, the control interface of the power generation equipment is responsible for receiving the control signal codes for this equipment and converting them into actual operation instructions to adjust the operation parameters of the equipment.

[0111] When transmitting the control signal code for the power generation equipment, such as the control signal code frame for adjusting the excitation current of the generator generated previously, to the control interface of the power generation equipment, the interface first verifies the code. It calculates the check bits in the control signal code frame according to the verification method specified in the communication protocol, such as CRC (Cyclic Redundancy Check), and compares them with the received check bits. If the calculated check bits are the same as the received ones, it is considered that the control signal code frame has not erred during transmission, and subsequent processing continues; if they are inconsistent, the code frame is discarded, and an error message is sent to the sending end, requesting retransmission.

[0112] After the verification passes, the control interface analyzes the information in the control signal code frame. It determines the type of operation to be performed according to the function code. For example, a function code of Func1 indicates adjusting the excitation current. Then it extracts the register address Reg1 and the encoded data content from the code frame, and restores the data content to the actual parameter value, i.e., the value I1, according to the decoding method specified in the protocol. Next, based on the information parsed out, the control interface generates corresponding electrical signals or control instructions through internal circuits or software logic, and sends them to the relevant control units of the generator, thereby triggering the adjustment operation of the generator excitation current and adjusting the excitation current to the value I1, achieving precise adjustment of the operating parameters of the power generation equipment.

[0113] For other equipment such as water turbines, it also follows a similar process. The corresponding control signal code is injected into its control interface, and after verification and analysis, it triggers the corresponding operating parameter adjustment operation, ensuring that various operating equipment in the hydropower station can accurately adjust parameters according to the instructions in the equipment operation and maintenance optimization strategy set to cope with abnormal situations of the equipment, restore the normal operating state of the equipment or ensure the stable operation of the equipment under abnormal conditions.

[0114] Step S153: Synchronously analyze the maintenance resource scheduling plan to generate a maintenance task work order and a resource allocation instruction.

[0115] The maintenance resource scheduling plan contains important information such as a spare parts requirement list, a personnel scheduling path, and maintenance time window allocation parameters. It is analyzed to generate specific maintenance task work orders and resource allocation instructions.

[0116] First, look at the spare parts requirement list. Assume the list content is [(Sp1, N1, R1), (Sp2, N2, R2), (Sp3, N3, R3)], where (Sp1, N1, R1) represents spare part type Sp1, quantity N1, and emergency allocation path R1. Analyze this list and generate corresponding resource allocation instructions for each spare part. For spare part Sp1, the resource allocation instruction may include details such as extracting N1 Sp1 spare parts from the warehouse and transporting them according to the allocation path R1. For example, the instruction content is "Extract N1 Sp1 spare parts from warehouse W1 (assuming Sp1 is stored in warehouse W1), and hand them over to the transportation team to transport them to the target component installation location according to route R1". Similarly, generate similar detailed resource allocation instructions for spare parts Sp2 and Sp3.

[0117] Next, based on the personnel scheduling path information, such as [(P1, Y1, R4, t2), (P3, Y2, R5, t3), (P4, Z1, R6, t4)], where (P1, Y1, R4, t2) means maintenance personnel P1 goes to equipment Y1, via route R4, and the expected arrival time is t2. Generate the personnel allocation part of the maintenance task work order. For maintenance personnel P1, record in the maintenance task work order "Maintenance personnel P1 goes to equipment Y1 via route R4 at time t2 to perform the maintenance task for Y1". At the same time, combine the maintenance time window allocation parameters, such as (T1, [tw1, tw2], R1, P5), which means that component T1 obtains spare parts through the spare part allocation path R1 within the maintenance time window [tw1, tw2], and is maintained by maintenance personnel P5. In the maintenance task work order, detail information such as the maintenance time window, required spare parts, and the personnel responsible for maintenance of each component. For example, the maintenance task work order for component T1 is recorded as "Within the time interval [tw1, tw2], maintenance personnel P5 uses the spare parts obtained through the allocation path R1 to maintain component T1, and the maintenance content is determined according to the equipment operation and maintenance optimization strategy (such as checking whether the component connections are loose, replacing worn parts, etc.)".

[0118] Through such a comprehensive analysis of the maintenance resource scheduling plan, generate detailed maintenance task work orders and resource allocation instructions, providing clear and definite guidance for the actual maintenance work, and ensuring that the maintenance work can be carried out orderly and efficiently.

[0119] Step S154: Distribute the maintenance task work order to the corresponding maintenance terminal device, and monitor the execution progress of the resource allocation instruction.

[0120] The maintenance terminal device can be a handheld device carried by maintenance personnel, a computer terminal installed at the maintenance site, etc. When the maintenance task work order is generated, it needs to be accurately distributed to the corresponding maintenance terminal device so that maintenance personnel can obtain task information in a timely manner and carry out their work.

[0121] Suppose maintenance personnel P1 is responsible for the maintenance task of equipment Y1, and the corresponding maintenance terminal device is terminal T1. Through the communication network within the hydropower station, a maintenance task work order containing the detailed information of P1 going to equipment Y1 to perform the maintenance task is sent to terminal T1. The communication network may use a wired network (such as Ethernet) or a wireless network (such as Wi-Fi, 4G / 5G, etc.) for data transmission. During the sending process, in order to ensure the accuracy and integrity of the data, the maintenance task work order will be encapsulated and verified. For example, the TCP / IP protocol is used for encapsulation, adding header and trailer information. The header contains information such as the address of the target terminal device (i.e., the address of terminal T1) and the data length, and the trailer contains information such as the checksum for verifying the data integrity. After receiving the maintenance task work order, terminal T1 first verifies it. If the verification passes, the work order information is displayed on the terminal interface, prompting maintenance personnel P1 that there is a new maintenance task.

[0122] At the same time, monitor the execution progress of the resource allocation instruction. Taking the allocation instruction of spare part Sp1 "Extract N1 Sp1 spare parts from warehouse W1 and hand them over to the transportation team to transport to the target component installation location according to route R1" as an example, set sensors in warehouse W1 or use the inventory management system to record the extraction time and quantity of the spare parts. When N1 Sp1 spare parts are extracted from the warehouse, the system records the extraction time and feeds this information back to the monitoring center. Along transportation route R1, multiple monitoring points may be set, such as GPS positioning devices installed on the transportation vehicle and logistics monitoring stations along the way. These monitoring points collect the location information and transportation status information of the transportation vehicle in real time (such as whether it is driving normally, whether it encounters a fault, etc.) and transmit this information to the monitoring center. The monitoring center updates the allocation progress of spare part Sp1 in real time based on this feedback information. For example, it displays on the monitoring system interface that "Spare part Sp1 has been extracted from the warehouse, the transportation vehicle has traveled to the XX location on route R1, and it is expected to arrive at the target component installation location at XX time". For all resource allocation instructions, they are monitored in this way in real time to promptly discover possible problems (such as transportation delays, spare part shortages, etc.) and take corresponding measures for adjustment to ensure that the maintenance resources can reach the maintenance site on time and accurately, guaranteeing the smooth progress of the maintenance work.

[0123] Step S155: Within the monitoring time window, collect the adjusted status data and maintenance operation feedback data of the device through the real-time data stream collection device, update the real-time monitoring data set in real time based on the event trigger mechanism, and dynamically optimize the response frequency of the closed-loop operation and maintenance management process through the sliding window mechanism.

[0124] Within each monitoring time window, the real-time data stream acquisition device of the hydropower station continuously operates to collect the status data after equipment adjustment and the maintenance operation feedback data. Taking the power generation equipment as an example, the real-time data stream acquisition device may include various sensors. For instance, after adjusting the excitation current of the generator, the current sensor will continuously monitor the actual value of the excitation current in real time, and the temperature sensor will monitor the temperature change during the operation of the generator, etc. The data collected by these sensors is transmitted to the data processing center in the form of a data stream.

[0125] The maintenance operation feedback data comes from the data uploaded by the maintenance personnel through the maintenance terminal device during the execution of maintenance tasks. For example, after maintenance personnel P1 performs maintenance on equipment Y1, they upload the maintenance result information through terminal T1, such as "Equipment Y1 has completed maintenance, and after inspection, the operating parameters have returned to the normal range", or "During the maintenance of equipment Y1, it is found that there are still potential problems with XX components and further attention is required", etc.

[0126] Based on the event-triggering mechanism, the real-time monitoring data set is updated in real time. Suppose when the current sensor monitors that the excitation current of the generator exceeds the predetermined normal range, this abnormal event triggers the data update operation. After receiving the information of this abnormal event, the data processing center immediately adds the status data of the generator at the current moment (including the abnormal excitation current value, temperature and other relevant data) and the maintenance operation feedback data (if any) to the real-time monitoring data set. In this way, the real-time monitoring data set can timely reflect the latest status of the equipment after operation and maintenance operations, providing accurate data support for subsequent analysis and decision-making.

[0127] The response frequency of the closed-loop operation and maintenance management process is dynamically optimized through the sliding window mechanism. The sliding window mechanism sets a time window of fixed length, for example, the time window length is ΔT. In each time window, the data in the real-time monitoring data set is analyzed. Assuming that in the current time window, through the analysis of the equipment status data, it is found that the equipment abnormalities have increased, or the maintenance operation feedback data shows that the maintenance effect is not ideal, according to the preset rules, the length of the sliding window is shortened, for example, ΔT is shortened to ΔT' (ΔT'<ΔT). In this way, in the next time window, the data can be analyzed more frequently, the slight changes in the equipment status can be captured in time, and potential problems can be discovered more quickly, thereby improving the response frequency of the closed-loop operation and maintenance management process, enabling the operation and maintenance personnel to take measures to deal with equipment abnormalities more promptly, and further optimize the operation and maintenance management effect. On the contrary, if in the current time window, the equipment status is stable and the maintenance operation feedback is good, the length of the sliding window can be appropriately extended according to the rules, the frequency of data analysis can be reduced, and the consumption of system resources can be reduced, while maintaining effective monitoring of the equipment status. By dynamically adjusting the sliding window length, the response frequency of the closed-loop operation and maintenance management process can be optimized, ensuring that the hydropower station equipment is always under efficient and stable operation and maintenance management.

[0128] It is worth noting that in the above embodiments, those skilled in the art can consider the dimensional uniformity and characteristic dimension matching of different types of data based on the knowledge of relevant technologies. For various types of data in the real-time monitoring data set, such as current, temperature, etc., they have different dimensions. When performing data analysis and processing, normalization or standardization processing is required for data that need to be calculated or compared with each other. For example, for current data I and temperature data T, if these two factors need to be considered comprehensively in a certain analysis model, the current data I is first standardized, assuming that its value range is [min_I, max_I], and it is converted into dimensionless standardized data I' by formula I'=(I-min_I) / (max_I-min_I). The temperature data T is also processed similarly, assuming that its value range is [min_T, max_T], and the standardized data T' is obtained by formula T'=(T-min_T) / (max_T-min_T). In this way, in subsequent analysis and calculation, the problem of inconsistent dimensions is avoided, ensuring the accuracy and reliability of the analysis results. At the same time, when processing multi-dimensional data, such as the equipment status feature vector that may contain features of multiple dimensions, it is necessary to ensure the matching of feature dimensions in various calculations and operations. For example, when performing feature fusion or weighted calculation, ensure that the dimensions of each feature vector involved in the calculation are consistent. If they are inconsistent, adjust them through appropriate methods, such as padding or dimensionality reduction, to meet the calculation requirements and ensure the scientificity and rationality of data processing in the entire operation and maintenance management process.

[0129] Figure 2 FIG. shows a schematic diagram of exemplary hardware and software components of an operation and maintenance management system 100 for a hydropower station that can implement the inventive concept provided by some embodiments of the present invention. For example, a processor 120 can be used on the operation and maintenance management system 100 of the hydropower station and is used to execute the functions in the present invention.

[0130] The operation and maintenance management system 100 for a hydropower station can be a general-purpose server or a special-purpose server, both of which can be used to implement the operation and maintenance management method for a hydropower station of the present invention. Although only one server is shown in the present invention, for convenience, the functions described in the present invention can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0131] For example, the operation and maintenance management system 100 for a hydropower station can include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as disks, ROM, or RAM, or any combination thereof. Exemplarily, the operation and maintenance management system 100 for a hydropower station can 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 can be implemented according to these program instructions. The operation and maintenance management system 100 for a hydropower station also includes an I / O interface 150 between the computer and other input / output devices.

[0132] For ease of explanation, only one processor is described in the operation and maintenance management system 100 for a hydropower station. However, it should be noted that the operation and maintenance management system 100 in the present invention can also include multiple processors. Therefore, the steps executed by one processor described in the present invention can also be jointly executed or separately executed by multiple processors. For example, if the processor of the operation and maintenance management system 100 for a hydropower station executes step A and step B, it should be understood that step A and step B can also be jointly executed by two different processors or separately executed in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor jointly execute steps A and B.

[0133] In addition, an embodiment of the present invention also provides a readable storage medium, in which computer-executable instructions are preset. When the processor executes the computer-executable instructions, the above-mentioned operation and maintenance management method for a hydropower station is implemented.

[0134] It should be noted that, in order to simplify the presentation of the disclosure of the present invention and thus help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, sometimes multiple features are merged into one embodiment, drawing, or description thereof.

Claims

1. An operation and maintenance management method for a hydropower station, characterized in that, The method includes: Obtaining a real-time monitoring data set of the operating equipment of the hydropower station, where the real-time monitoring data set contains sensor data streams of multiple monitoring time windows; Performing a dynamic feature extraction operation on the real-time monitoring data set to generate a state feature set of the operating equipment of the hydropower station, where the state feature set includes equipment operation stability features, environmental coupling features, and abnormal fluctuation correlation features; Based on a preset abnormal detection strategy network, performing an abnormal state matching process on the state feature set to determine the abnormal level and abnormal influence range parameters of the operating equipment of the hydropower station; Generating an equipment operation and maintenance optimization strategy set according to the abnormal level and abnormal influence range parameters, where the equipment operation and maintenance optimization strategy set contains an equipment control instruction sequence and a maintenance resource scheduling plan; Feeding back the equipment operation and maintenance optimization strategy set to the hydropower station control center to trigger equipment operation and maintenance operations; The performing a dynamic feature extraction operation on the real-time monitoring data set to generate a state feature set of the operating equipment of the hydropower station includes: For the sensor data stream of each monitoring time window, performing time-domain segmentation processing to obtain multiple data segment sets; Invoking a preset feature encoding model to perform multi-scale feature fusion processing on each data segment set to generate a time-frequency joint feature vector of the data segment set; Constructing equipment operation stability features based on the time-frequency joint feature vector, where the equipment operation stability features include vibration amplitude change trend parameters, temperature fluctuation correlation parameters, and pressure gradient matching parameters, and the vibration amplitude change trend parameters are generated based on the peak sequence of the time-domain vibration signal and the frequency-domain energy integral value; Extracting the difference information of the sensor data streams between adjacent monitoring time windows to generate environmental coupling features, where the environmental coupling features include water flow velocity change influence factors, water level dynamic correlation factors, and external environment disturbance coefficients; Performing abnormal fluctuation pattern recognition processing on the sensor data stream to generate abnormal fluctuation correlation features, where the abnormal fluctuation correlation features include sudden fluctuation frequencies, fluctuation duration correlation parameters, and fluctuation propagation path topological relationships; Performing standardization processing on the equipment operation stability features, the environmental coupling features, and the abnormal fluctuation correlation features respectively, and then merging them into the state feature set.

2. The operation and maintenance management method for a hydropower station according to claim 1, characterized in that The invoking a preset feature encoding model to perform multi-scale feature fusion processing on each data segment set to generate a time-frequency joint feature vector of the data segment set includes: Performing fast Fourier transform processing on each data segment set to generate a frequency-domain energy distribution matrix; Extracting the energy peak sequence and energy mean parameter within a preset frequency band range in the frequency-domain energy distribution matrix; Performing time-domain difference calculation processing on the data segment set to generate a time-domain change rate sequence and time-domain mutation point position information; Performing standardization conversion on the energy peak sequence, the energy mean parameter, the time-domain change rate sequence, and the time-domain mutation point position information respectively, and then inputting them into the feature encoding model for feature cross-fusion processing to generate a fusion feature vector containing time-domain correlation weights and frequency-domain correlation weights; Generate the time-frequency joint feature vector based on the similarity calculation between the fused feature vector and a preset reference feature template.

3. The operation and maintenance management method for a hydropower station according to claim 1, characterized in that The extraction of the difference information of the sensor data stream between adjacent monitoring time windows to generate an environmental coupling feature includes: Obtain the difference data distribution between the sensor data stream of the current monitoring time window and the sensor data stream of the previous monitoring time window; Perform clustering analysis on the difference data distribution to determine the difference clustering center point and the clustering radius parameter; After uniformly converting the unit of the difference clustering center point and matching and calculating it with the unit in the preset environmental impact factor mapping table, generate a water flow velocity change impact factor; Conduct a correlation analysis based on the clustering radius parameter and the water level fluctuation range in the historical environmental data to generate a water level dynamic correlation factor; Extract the timestamp of the sudden disturbance event and the disturbance intensity parameter in the difference data distribution, and combine with the external environmental monitoring data to generate an external environmental disturbance coefficient; Combine the water flow velocity change impact factor, the water level dynamic correlation factor, and the external environmental disturbance coefficient as the environmental coupling feature.

4. The operation and maintenance management method for a hydropower station according to claim 1, characterized in that, Based on a preset anomaly detection strategy network, perform anomaly state matching processing on the state feature set to determine the anomaly level and the anomaly impact range parameter of the hydropower station operation equipment, including: Input the state feature set into the anomaly detection strategy network for feature importance ranking processing to generate a feature weight distribution vector; Perform weighted fusion processing on the state feature set based on the feature weight distribution vector to generate a comprehensive equipment state score; Dynamically generate a dynamic anomaly threshold interval adaptively adjusted based on the equipment operation state according to the type of the hydropower station operation equipment and the historical normal operation data, compare the comprehensive equipment state score with the dynamic anomaly threshold interval, and if the comprehensive equipment state score exceeds the dynamic anomaly threshold interval, trigger an anomaly level classification process; Call a preset anomaly impact assessment model, and calculate the anomaly impact range parameter according to the anomaly fluctuation correlation feature and the environmental coupling feature in the state feature set. The anomaly impact range parameter includes equipment component association degree parameter, upstream and downstream equipment impact coefficient, and maintenance priority score; Determine the anomaly level according to the deviation degree between the comprehensive equipment state score and the anomaly threshold interval and the anomaly impact range parameter.

5. The operation and maintenance management method for a hydropower station according to claim 4, characterized in that The call of the preset anomaly impact assessment model to calculate the anomaly impact range parameter according to the anomaly fluctuation correlation feature and the environmental coupling feature in the state feature set includes: Extract the topological relationship of the fluctuation propagation path in the anomaly fluctuation correlation feature to construct an equipment component association graph; Use a node centrality algorithm based on betweenness centrality to process the equipment component association graph to determine the core component nodes and edge component nodes of the hydropower station operation equipment; Generate an equipment component association degree parameter according to the historical maintenance record and the current state parameter of the core component node. Combined with the water flow velocity change influencing factor and the water level dynamic correlation factor in the environmental coupling characteristics, calculate the hydraulic coupling coefficient between the operating equipment of the hydropower station and the upstream and downstream equipment; Based on the hydraulic coupling coefficient and the state parameters of the edge component nodes, generate the influence coefficients of the upstream and downstream equipment; According to the equipment component correlation degree parameter and the influence coefficients of the upstream and downstream equipment, generate a maintenance priority score through a weighted evaluation algorithm; Combine the equipment component correlation degree parameter, the influence coefficients of the upstream and downstream equipment, and the maintenance priority score into the abnormal influence range parameter; 6. The operation and maintenance management method for a hydropower station according to claim 1, characterized in that Generating an equipment operation and maintenance optimization strategy set according to the abnormal level and the abnormal influence range parameter, including: Match the preset operation and maintenance response strategy template according to the abnormal level to generate a basic control instruction sequence; Based on the maintenance priority score in the abnormal influence range parameter, perform dynamic adjustment processing on the basic control instruction sequence to generate an optimized equipment control instruction sequence; Extract the equipment component correlation degree parameter and the influence coefficients of the upstream and downstream equipment in the abnormal influence range parameter, and generate a maintenance resource scheduling plan in combination with the maintenance resource inventory data. The maintenance resource scheduling plan includes a spare part requirement list, a personnel scheduling path, and a maintenance time window allocation parameter; Perform time-axis alignment processing on the equipment control instruction sequence and the maintenance resource scheduling plan to maintain the timing consistency between the execution of the equipment control instruction and the resource scheduling operation; Combine the aligned equipment control instruction sequence and the maintenance resource scheduling plan into the equipment operation and maintenance optimization strategy set; 7. The operation and maintenance management method for a hydropower station according to claim 6, characterized in that The performing dynamic adjustment processing on the basic control instruction sequence based on the maintenance priority score in the abnormal influence range parameter to generate an optimized equipment control instruction sequence includes: Obtain the initial instruction execution time point and the instruction parameter range in the basic control instruction sequence; Determine the instruction urgency coefficient according to the maintenance priority score, and dynamically adjust the time window length of the initial instruction execution time point based on the collaborative optimization result of the instruction urgency coefficient and the maintenance time window allocation parameter; Adjust the control intensity parameter and the control duration parameter in the instruction parameter range in combination with the equipment component correlation degree parameter; Based on the influence coefficients of the upstream and downstream equipment, insert upstream and downstream equipment collaborative control instructions into the basic control instruction sequence; Combine the adjusted instruction execution time point, the instruction parameter range, and the collaborative control instructions into the optimized equipment control instruction sequence; 8. The operation and maintenance management method for a hydropower station according to claim 6, characterized in that The extracting the equipment component correlation degree parameter and the influence coefficients of the upstream and downstream equipment in the abnormal influence range parameter and generating a maintenance resource scheduling plan in combination with the maintenance resource inventory data includes: Screen the target component set that needs to be preferentially maintained according to the equipment component correlation degree parameter; Based on the historical failure data, the current state parameters of the target component set, and the fluctuation propagation path topology relationship in the abnormal fluctuation correlation characteristics, dynamically generate a spare part requirement list. The spare part requirement list includes spare part types, quantities, and emergency allocation paths; Determine the list of affected external devices according to the upstream and downstream device impact coefficients, and generate a personnel scheduling path in combination with the maintenance personnel location data. The personnel scheduling path includes path priority and time estimation parameters; Based on the maintenance time window allocation parameters, perform space-time conflict detection processing on the spare part allocation path and the personnel scheduling path. If a space-time conflict is detected, dynamically adjust the spare part allocation path or the personnel scheduling path according to the maintenance priority score to generate maintenance time window allocation parameters without space-time conflicts; Integrate the spare part demand list, the personnel scheduling path, and the maintenance time window allocation parameters into the maintenance resource scheduling plan.

9. An operation and maintenance management system for a hydropower 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 operation and maintenance management method for hydropower stations described in any one of the above claims 1-8.

Images