Power monitoring method, system and equipment based on marine thermal anomaly influence

Abstract

The invention relates to the field of electric power monitoring, in particular to an electric power monitoring method, system and device based on ocean thermal anomaly influence. According to the method, through an innovative ocean thermal anomaly analysis algorithm, ocean thermal anomaly features are accurately extracted, a key correction basis is provided for power load prediction, and the prediction precision is improved; meanwhile, a scientific and reasonable power load prediction model is established, and the comprehensive influence of ocean thermal abnormal factors and the atmospheric environment and power system data is fully considered; besides, an early warning threshold value based on multi-factor comprehensive consideration is set, and the accuracy and timeliness of monitoring and early warning are improved, so that stable operation of a power system under the influence of marine thermal anomaly is guaranteed, and power failures and losses caused by marine environment changes are reduced.

Description

Technical Field

[0001] The present application relates to the field of power monitoring, and in particular to a power monitoring method, system and equipment based on the influence of ocean thermal anomalies. Background Art

[0002] In the field of power system operation today, power load forecasting plays a vital role in ensuring the stability and reliability of power supply. Traditional power load forecasting methods mainly focus on traditional meteorological factors (such as atmospheric temperature, atmospheric pressure, etc.) and historical data of the power system itself. However, with the in-depth study of the interaction between the ocean and the atmosphere, people gradually realize that ocean thermal anomalies have an impact on the power system that cannot be ignored.

[0003] Ocean thermal anomalies can indirectly cause significant fluctuations in power load by changing complex physical processes such as atmospheric circulation. However, in the process of power load forecasting, existing technologies only rely on traditional meteorological factors and the power system's own historical data, and lack effective consideration of the deep and complex relationship between ocean thermal anomalies and power load. This limitation directly leads to the inability to accurately capture the characteristics of power load changes caused by ocean thermal anomalies, which in turn causes the accuracy of power load forecasting to be at a low level.

[0004] When an abnormal event occurs in the ocean, it is difficult to timely and effectively adjust the power load forecast results according to the dynamic changes of the marine environment because the existing forecast model does not fully integrate the ocean-related data and its association mechanism with the power load. As a result, there is often a large degree of deviation between power supply and actual demand. This deviation may cause unnecessary waste of energy on the one hand, and may also cause power shortage on the other hand, seriously affecting the normal operation of the power system and the user's power experience.

[0005] In addition, the existing technology is too simple or even ignores the processing of ocean-related data in dealing with the impact of ocean thermal anomalies on power systems. This not only makes it impossible to effectively warn of the risks that ocean thermal anomalies may bring to the power system in advance, but also makes it difficult to provide comprehensive, in-depth and reliable decision-making basis for the operation and maintenance personnel of the power system. In terms of setting warning thresholds, the existing methods are not scientific and reasonable, and false alarms or missed alarms often occur, which undoubtedly greatly reduces the normal operation efficiency and maintenance efficiency of the power system, and is not conducive to the power system in the complex and changeable marine environment and the background of global climate change to achieve safe, stable and efficient operation goals.

[0006] In summary, there is an urgent need for a new technical solution to overcome the shortcomings of existing technologies, to fully consider the impact of ocean thermal anomalies on the power system, to improve the accuracy of power load forecasting, and to achieve scientific and reasonable early warning to ensure the reliable operation of the power system. Summary of the invention

[0007] In view of this, an object of the present invention is to provide a method, system and device for monitoring power based on the influence of ocean thermal anomalies, so as to solve at least one problem existing in the prior art.

[0008] According to a first aspect of an embodiment of the present invention, there is provided a method for monitoring power based on the influence of ocean thermal anomalies, the method comprising:

[0009] Use multiple sensors to collect multi-source data; the multi-source data includes: ocean temperature data, ocean current data, atmospheric temperature data, atmospheric pressure data, voltage, current, power data of each substation, and load data of power lines;

[0010] Using the ocean temperature data and the ocean current data, a composite sequence is constructed, and the composite sequence is subjected to spectral decomposition to obtain all detail components and approximate components, and all detail components and approximate components are used to perform feature reconstruction, and the result of the feature reconstruction, preset singular value decomposition and index construction are used to obtain an ocean thermal anomaly feature index;

[0011] Using the ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, a basic prediction model input vector is constructed, and the basic prediction model input vector is corrected according to the ocean thermal anomaly state to obtain a corrected input vector;

[0012] Using the corrected input vector and the prediction model based on the long short-term memory network, a power load prediction model is constructed;

[0013] Using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data of the period of ocean thermal anomaly, the power load prediction model is trained using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load;

[0014] Using the prediction result of the power load and the ocean thermal anomaly characteristic index, warning thresholds are set, including an overload warning threshold and a temperature anomaly warning threshold.

[0015] Furthermore, the ocean temperature data and the ocean current data are used to construct a composite sequence, and the composite sequence is subjected to spectral decomposition to obtain all detail components and approximate components, including:

[0016] S1. Using the ADTF algorithm, define the initial decomposition components:

[0017] S2. At each decomposition, the instantaneous frequency is calculated based on the Hilbert transform of the current decomposition component;

[0018] S3. Constructing a time-frequency filtering window according to the instantaneous frequency and the empirical coefficient;

[0019] S4. Filter the current decomposition component using the time-frequency filter window to obtain detail components and approximate components;

[0020] S5. Repeat the decomposition steps S2-S4 until the decomposition times reach the upper limit, and obtain all detail components and approximate components.

[0021] Furthermore, the feature reconstruction using all detail components and approximate components includes:

[0022] Normalize each detail component into a specific form by calculating its mean and standard deviation;

[0023] Normalize each approximate component into a specific form by calculating its mean and standard deviation;

[0024] Constructing a feature matrix using the normalized values ​​of the detail components at different time points as row vectors;

[0025] The normalized approximate component is added as the last row vector of the feature matrix to obtain a new feature matrix.

[0026] Furthermore, the result of the feature reconstruction, the preset singular value decomposition and the index construction are used to obtain the ocean thermal anomaly feature index, including:

[0027] Perform singular value decomposition on the new feature matrix to obtain a singular value matrix;

[0028] Selecting left singular vectors corresponding to a specific number of singular values ​​in the singular value matrix for feature reconstruction;

[0029] Elements are selected from the reconstructed eigenvector and combined with the corresponding weight coefficients to construct the ocean thermal anomaly characteristic index.

[0030] Furthermore, the use of the ocean thermal anomaly state, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data to construct a basic prediction model input vector, and the basic prediction model input vector is corrected according to the ocean thermal anomaly characteristic index to obtain a corrected input vector, including:

[0031] The ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data are obtained, and the lag order is determined according to the autocorrelation analysis of the historical power load data to construct the basic prediction model input vector, as shown in the following formula:

[0032] X(t)=[L(t-1),L(t-2),…,L(tn),A(t),P(t)] (1)

[0033] Where n is the lag order determined based on the autocorrelation analysis of historical power load data, A(t) represents the atmospheric temperature data, P(t) represents the atmospheric pressure data, and L(t) represents the power system operation data;

[0034] According to the ocean thermal anomaly state, when the preset conditions are met, the ocean thermal anomaly correction term is introduced into the basic prediction model input vector to obtain the corrected input vector, as shown in the following formula:

[0035] X m (t)=[L(t-1),L(t-2),…,L(tn),A(t),P(t),γF(t)] (2)

[0036] in, I represents the characteristic index of ocean thermal anomaly, and γ is the weight coefficient of the impact of ocean thermal anomaly on power load.

[0037] Furthermore, the use of the corrected input vector and the prediction model based on the long short-term memory network to construct a power load prediction model includes:

[0038] Built on LSTM network, including input layer, hidden layer and output layer;

[0039] The number of neurons in the input layer is based on the corrected input vector X m The dimension of (t) is determined by X m The number of elements in (t) is the same;

[0040] The hidden layer is set to a multi-layer structure and the number of layers is determined by the mean square error as the evaluation index. Each LSTM unit in the hidden layer contains an input gate i t , Forget gate t , output gate o t and memory unit c t , hidden state h t , and calculated according to the following formula:

[0041]

[0042]

[0043] where x t The input vector X m (t) value at time t, h t-1 is the hidden state at the previous moment, W i ,W f ,Wo ,W c are the weight matrices corresponding to the gate and memory unit, respectively, i ,b f ,b o ,b c is the bias vector, ⊙ represents element-by-element multiplication, σ is the sigmoid function, and the number of neurons in the output layer is 1, corresponding to the predicted power load value.

[0044] Furthermore, the power load data, the atmospheric temperature data, the atmospheric pressure data and the relevant data of the period of ocean thermal anomaly are used to train the power load prediction model using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load, including:

[0045] The power load prediction model is trained using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data during the period of ocean thermal anomaly, and the connection weights of the neural network are optimized using an adaptive moment estimation optimization algorithm to update the network parameters;

[0046] Using the preset total number of training steps and the current number of training steps, the learning rate is adjusted through a dynamic learning rate adjustment strategy to obtain the prediction result of the power load;

[0047] The power load prediction model is trained by using the power load data, the atmospheric temperature data, the atmospheric pressure data and the relevant data of the ocean thermal anomaly period, and the connection weights of the neural network are optimized by using the adaptive moment estimation optimization algorithm to update the network parameters, including:

[0048] Initialize the first-order moment vector and the second-order moment vector to 0, and set the learning rate to the initial value η0;

[0049] The gradient is calculated according to the preset loss function; the loss function is as follows:

[0050]

[0051] Among them, y i is the true value, is the predicted value, θ is the network parameter;

[0052] The first-order moment vector is updated using a preset first formula; wherein the preset first formula is as follows:

[0053] m t =β1m t-1 +(1-β1)g t (5)

[0054] Where β1 is a hyperparameter used to control the moving average weight of the gradient;

[0055] The second-order moment vector is updated using a preset second formula; wherein the preset second formula is as follows:

[0056]

[0057] Where β2 is a hyperparameter used to control the moving average weight of the squared gradient;

[0058] respectively performing deviation correction on the first-order moment vector and the second-order moment vector;

[0059] The network parameters are updated according to the first-order moment vector and the second-order moment vector after the deviation correction.

[0060] Furthermore, the method uses the preset total number of training steps and the current number of training steps to adjust the learning rate through a dynamic learning rate adjustment strategy to obtain a prediction result of the power load, including:

[0061] Using the preset total number of training steps and the current number of training steps, as shown in the following formula, the learning rate is adjusted through the dynamic learning rate adjustment strategy to obtain the prediction result of the power load;

[0062] In the early stage of training When the learning rate η t =η0;

[0063] In the middle of training hour,

[0064] In the later stages of training hour,

[0065] Where N is the total number of training steps, t is the current number of training steps, η 0为 Initial learning rate.

[0066] According to a second aspect of an embodiment of the present invention, a power monitoring system based on the influence of ocean thermal anomalies is provided, which is applied to any of the above-mentioned power monitoring methods based on the influence of ocean thermal anomalies, and the system includes:

[0067] The data acquisition module is used to collect multi-source data using a variety of sensors; the multi-source data includes: ocean temperature data, ocean current data, atmospheric temperature data, atmospheric pressure data, voltage, current, power data of each substation, and load data of the power line;

[0068] The ocean thermal anomaly analysis module is used to construct a composite sequence using the ocean temperature data and the ocean current data, perform spectrum decomposition on the composite sequence to obtain all detail components and approximate components, perform feature reconstruction using all detail components and approximate components, and obtain ocean thermal anomaly feature indicators using the result of the feature reconstruction, preset singular value decomposition and indicator construction;

[0069] The power load forecasting module is used to construct a basic forecasting model input vector by using the ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, and to correct the basic forecasting model input vector according to the ocean thermal anomaly state to obtain a corrected input vector;

[0070] Using the corrected input vector and the prediction model based on the long short-term memory network, a power load prediction model is constructed;

[0071] Using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data of the period of ocean thermal anomaly, the power load prediction model is trained using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load;

[0072] The monitoring and early warning module is used to use the prediction results of the power load and the ocean thermal anomaly characteristic indicators to set early warning thresholds, including overload early warning thresholds and temperature anomaly early warning thresholds.

[0073] According to a third aspect of an embodiment of the present invention, there is provided a power monitoring device based on the influence of ocean thermal anomalies, the device comprising:

[0074] A memory having an executable program stored therein;

[0075] A processor is used to execute the executable program in the memory to implement the steps of any one of the methods described above.

[0076] The technical solution provided by the embodiments of the present invention may have the following beneficial effects:

[0077] It can be understood that the technical solution provided by the present invention uses multiple sensors to collect multi-source data;

[0078] A composite sequence is constructed by using the ocean temperature data and the ocean current data, and the composite sequence is spectrally decomposed to obtain all detail components and approximate components. All detail components and approximate components are used to perform feature reconstruction, and the result of the feature reconstruction, the preset singular value decomposition and the index construction are used to obtain the characteristic index of ocean thermal anomaly. Afterwards, a basic prediction model input vector is constructed by using the characteristic index of ocean thermal anomaly, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, and the basic prediction model input vector is corrected according to the state of ocean thermal anomaly to obtain the corrected input vector. A power load prediction model is constructed by using the corrected input vector and a prediction model based on a long short-term memory network. The power load prediction model is trained by using the power load data, the atmospheric temperature data, the atmospheric pressure data, and the relevant data of the period of ocean thermal anomaly, and a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain the prediction result of the power load. Finally, the prediction result of the power load and the characteristic index of ocean thermal anomaly are used to set warning thresholds, including overload warning thresholds and temperature anomaly warning thresholds.

[0079] It is understandable that the technical solution provided by the present invention, through the ocean thermal anomaly analysis module, deeply explores the intrinsic connection between ocean temperature and ocean current data, extracts the characteristics of ocean thermal anomalies, and provides a key correction basis for power load forecasting. When ocean thermal anomalies occur, it can timely capture the potential fluctuation trend of power load caused by changes in the ocean environment, avoid prediction deviations caused by ignoring ocean factors, effectively balance power supply and demand, reduce energy waste and power shortages caused by inaccurate load forecasts, and ensure stable operation of the power system. Based on the overload warning threshold and the temperature anomaly warning threshold, it can fully consider the impact of ocean thermal anomalies on various links of the power system, greatly improve the accuracy of the warning, and when the power system faces potential risks, it can timely issue warning signals and provide detailed response strategy suggestions, shorten the troubleshooting and repair time, and improve the operation and maintenance efficiency of the power system.

[0080] Through innovative ocean thermal anomaly analysis algorithms, the characteristics of ocean thermal anomalies can be accurately extracted to provide key correction basis for power load forecasting and improve prediction accuracy. At the same time, a scientific and reasonable power load forecasting model is established to fully consider the combined influence of ocean thermal anomaly factors and the atmospheric environment and the power system's own data. In addition, an early warning threshold based on comprehensive consideration of multiple factors is set to improve the accuracy and timeliness of monitoring and early warning, thereby ensuring the stable operation of the power system under the influence of ocean thermal anomalies and reducing power failures and losses caused by changes in the marine environment.

[0081] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0083] Figure 1 is a schematic diagram of steps of a power monitoring method based on the influence of ocean thermal anomalies according to an exemplary embodiment;

[0084] Figure 2 is a flow chart showing the composition of a power monitoring system based on the influence of ocean thermal anomalies according to an exemplary embodiment;

[0085] Figure 3 It is a schematic block diagram showing the composition of power monitoring equipment based on the influence of ocean thermal anomalies according to an exemplary embodiment. DETAILED DESCRIPTION

[0086] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0087] Embodiment 1

[0088] See also Figure 1 , Figure 1 The present invention is a schematic diagram of steps of a power monitoring method based on the influence of ocean thermal anomaly according to an exemplary embodiment, wherein the method comprises:

[0089] S1. Collect multi-source data using multiple sensors; the multi-source data includes: ocean temperature data, ocean current data, atmospheric temperature data, atmospheric pressure data, voltage, current, power data of each substation, and load data of power lines;

[0090] S2. construct a composite sequence using the ocean temperature data and the ocean current data, perform spectral decomposition on the composite sequence to obtain all detail components and approximate components, perform feature reconstruction using all detail components and approximate components, and obtain ocean thermal anomaly feature indicators using the result of the feature reconstruction, preset singular value decomposition and indicator construction;

[0091] S3. construct a basic prediction model input vector by using the ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, and modify the basic prediction model input vector according to the ocean thermal anomaly state to obtain a modified input vector;

[0092] S4. Using the corrected input vector and the prediction model based on the long short-term memory network, a power load prediction model is constructed;

[0093] S5. Using the power load data, the atmospheric temperature data, the atmospheric pressure data and the relevant data of the ocean thermal anomaly period, the power load forecasting model is trained using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a power load forecasting result;

[0094] S6. Using the predicted results of the power load and the ocean thermal anomaly characteristic indicators, set warning thresholds, including overload warning thresholds and temperature anomaly warning thresholds.

[0095] In one embodiment, the present application uses an innovative ocean thermal anomaly analysis algorithm to accurately extract the characteristics of ocean thermal anomalies, provide key correction basis for power load forecasting, and improve prediction accuracy; at the same time, establish a scientific and reasonable power load forecasting model, fully consider the combined influence of ocean thermal anomaly factors and the atmospheric environment and the power system's own data; in addition, set an early warning threshold based on comprehensive considerations of multiple factors to improve the accuracy and timeliness of monitoring and early warning, thereby ensuring the stable operation of the power system under the influence of ocean thermal anomalies and reducing power failures and losses caused by changes in the marine environment.

[0096] In one embodiment, as described in step S1, using multiple sensors to collect multi-source data includes:

[0097] A variety of sensors are used to collect multi-source data, including ocean temperature sensors, ocean current sensors, atmospheric temperature sensors, atmospheric pressure sensors, and power monitoring sensors installed on each substation and power lines. Ocean temperature sensors and ocean current sensors are deployed in relevant sea areas to collect ocean temperature data T(t) and ocean current data C(t). Atmospheric temperature sensors and atmospheric pressure sensors are distributed in multiple meteorological monitoring stations in the monitoring area to obtain atmospheric temperature data A(t) and atmospheric pressure data P(t) in real time. Power monitoring sensors continuously collect voltage, current, and power data from each substation and load data from power lines to form a power system operation data sequence L(t).

[0098] In one embodiment, as described in step S2, the ocean temperature data and the ocean current data are used to construct a composite sequence, and the composite sequence is subjected to spectral decomposition to obtain all detail components and approximate components, and all detail components and approximate components are used to perform feature reconstruction, and the result of the feature reconstruction, the preset singular value decomposition and the index construction are used to obtain the characteristic index of the ocean thermal anomaly, including:

[0099] After receiving the ocean temperature data T(t) and the ocean current data C(t), a composite sequence M(t) is constructed.

[0100] T(t)×ln(1+|C(t)|), perform spectral decomposition on M(t).

[0101] The specific steps of spectrum decomposition are as follows:

[0102] 1. Using the ADTF algorithm, define the initial decomposition component of M(t) as m0(t) = M(t).

[0103] 2. At the kth decomposition, calculate m k (t) is the instantaneous frequency Where H[m k (t)] is m k Hilbert transform of (t).

[0104] 3. According to the instantaneous frequency ω k (t) Constructing time-frequency filtering window in α and β are empirical coefficients.

[0105] 4. Use the time-frequency filter window W k (t,f) vs. m k (t) Filter to obtain detail components and approximate component a k (t) = m k (t)-d k (t).

[0106] 5. Repeat decomposition process 2-4 until the decomposition times reach the upper limit.

[0107] After the spectrum is decomposed through the above steps, all detail components and approximate components are used to perform feature reconstruction. The result of the feature reconstruction, the preset singular value decomposition and the index construction are used to obtain the characteristic index of the ocean thermal anomaly. The specific steps are as follows:

[0108] 1. Perform feature reconstruction on all the obtained detail components and approximate components, and convert the detail component d k (t) and approximate component a k(t) is decomposed into q detail components d1(t), d2(t),…,d q (t) and approximate component a q (t);

[0109] 2. For each detail component d i (t) is normalized to where μ i Yes i The mean of (t), σ i is its standard deviation, construct the feature matrix D, whose row vectors are the normalized detail components d i,n (t) values ​​at different time points, for the approximate component a q (t), and the same normalization is performed to obtain a q,n (t), and add it as the last row vector of the feature matrix D to get the new feature matrix D a ;

[0110] 3. For the feature matrix D a Perform singular value decomposition D a =UΣV T , select the left singular vectors u1,u2,…,u corresponding to the p singular values ​​in the singular value matrix Σ p , construct the reconstruction matrix U p =[u1u2…u p ];

[0111] 4. Feature reconstructed vector Select the elements in the reconstructed eigenvector to construct the ocean thermal anomaly characteristic index where λ i is the corresponding weight coefficient, the ocean thermal anomaly state is determined according to the value of I, and I is output to the power load prediction module.

[0112] In the specific implementation, as described in step S3, the ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data are used to construct a basic prediction model input vector, and the basic prediction model input vector is corrected according to the ocean thermal anomaly state to obtain a corrected input vector, including:

[0113] The ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data are obtained, and the lag order is determined according to the autocorrelation analysis of the historical power load data to construct the basic prediction model input vector, as shown in the following formula:

[0114] X(t)=[L(t-1),L(t-2),…,L(tn),A(t),P(t)] (1)

[0115] Where n is the lag order determined according to the autocorrelation analysis of historical power load data (the maximum lag order with a non-zero autocorrelation coefficient is selected as n by calculating the autocorrelation coefficient of the power load data at different lag orders), A(t) represents atmospheric temperature data, P(t) represents atmospheric pressure data, and L(t) represents power system operation data.

[0116] According to the ocean thermal anomaly state, when the preset conditions are met, the ocean thermal anomaly correction term is introduced into the basic prediction model input vector to obtain the corrected input vector, as shown in the following formula:

[0117] X m (t)=[L(t-1),L(t-2),…,L(tn),A(t),P(t),γF(t)] (2)

[0118] in, I represents the characteristic index of ocean thermal anomaly, and γ is the weight coefficient of the impact of ocean thermal anomaly on power load.

[0119] In the specific implementation, the basic prediction model is corrected according to the ocean thermal anomaly state. When S = 1, the ocean thermal anomaly correction term γF(t) is introduced into the input vector to obtain the corrected input vector X m (t).

[0120] In a specific implementation, the use of the corrected input vector and the prediction model based on the long short-term memory network to construct a power load prediction model includes:

[0121] Built on LSTM network, including input layer, hidden layer and output layer;

[0122] The number of neurons in the input layer is based on the corrected input vector X m The dimension of (t) is determined by X m The number of elements in (t) is the same;

[0123] The hidden layer is set to a multi-layer structure and the number of layers is determined by the mean square error as the evaluation index. Each LSTM unit in the hidden layer contains an input gate i t , Forget gate t , output gate o t and memory unit c t , hidden state h t , and calculated according to the following formula:

[0124]

[0125] where x t The input vector X m (t) value at time t, h t-1is the hidden state at the previous moment, W i ,W f ,W o ,W c are the weight matrices corresponding to the gate and memory unit, respectively, i ,b f ,b o ,b c is the bias vector, ⊙ represents element-by-element multiplication, σ is the sigmoid function, and the number of neurons in the output layer is 1, corresponding to the predicted power load value.

[0126] Furthermore, the power load data, the atmospheric temperature data, the atmospheric pressure data and the relevant data of the period of ocean thermal anomaly are used to train the power load prediction model using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load, including:

[0127] The power load prediction model is trained using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data during the period of ocean thermal anomaly, and the connection weights of the neural network are optimized using an adaptive moment estimation optimization algorithm to update the network parameters;

[0128] Using the preset total number of training steps and the current number of training steps, the learning rate is adjusted through a dynamic learning rate adjustment strategy to obtain the prediction result of the power load.

[0129] Furthermore, the power load prediction model is trained using the power load data, the atmospheric temperature data, the atmospheric pressure data, and the relevant data of the period of ocean thermal anomaly, and the connection weights of the neural network are optimized using an adaptive moment estimation optimization algorithm to update the network parameters, including:

[0130] Initialize the first-order moment vector m0 and the second-order moment vector v0 to 0, and set the learning rate η to the initial value η0;

[0131] The gradient is calculated according to the preset loss function; the loss function is as follows:

[0132]

[0133] Among them, y i is the true value, is the predicted value, θ is the network parameter;

[0134] The first-order moment vector is updated using a preset first formula; wherein the preset first formula is as follows:

[0135] m t =β1m t-1 +(1-β1)gt (5)

[0136] Where β1 is a hyperparameter used to control the moving average weight of the gradient;

[0137] The second-order moment vector is updated using a preset second formula; wherein the preset second formula is as follows:

[0138]

[0139] Where β2 is a hyperparameter used to control the moving average weight of the squared gradient;

[0140] respectively performing deviation correction on the first-order moment vector and the second-order moment vector;

[0141] The network parameters are updated according to the first-order moment vector and the second-order moment vector after the deviation correction.

[0142] Furthermore, the method uses the preset total number of training steps and the current number of training steps to adjust the learning rate through a dynamic learning rate adjustment strategy to obtain a prediction result of the power load, including:

[0143] Using the preset total number of training steps and the current number of training steps, as shown in the following formula, the learning rate is adjusted through the dynamic learning rate adjustment strategy to obtain the prediction result of the power load;

[0144] In the early stage of training When the learning rate η t =η0;

[0145] In the middle of training hour,

[0146] In the later stages of training hour,

[0147] Among them, N is the total number of training steps, t is the current number of training steps, and η0 is the initial learning rate.

[0148] In one embodiment, the LSTM network is trained by using historical power load data, atmospheric temperature data, atmospheric pressure data, and relevant data during the period of ocean thermal anomaly. During the training process, an adaptive moment estimation optimization algorithm is used to optimize the connection weights of the neural network. First, the first-order moment vector m0 and the second-order moment vector v0 are initialized to 0, and the learning rate η is set to the initial value η0. At the tth step of training, the gradient is calculated. Where J(θ) is the loss function, using the mean square error function y i is the true value, is the predicted value, θ is the network parameter, and the first-order moment vector m is updatedt =β1m t-1 +(1-β1)g t , the second-order moment vector Then the first-order moment vector and the second-order moment vector are corrected for deviation. Finally update the network parameters After each training cycle, the model performance is evaluated on the validation set. If the performance does not improve, the training is terminated early and the model parameters are saved. In addition, a dynamic learning rate adjustment strategy is adopted, that is, the total number of training steps is N, the current number of training steps is t, the initial learning rate is η0, and in the early stage of training When the learning rate η t =η0; in the middle of training hour, In the later stages of training hour,

[0149] More specifically, the present application receives the prediction result L of the power load prediction module. p (t k )(where t k represents the kth time point in the future) and the ocean thermal anomaly characteristic index I. Based on these data, the warning thresholds are set, including the overload warning threshold T o and temperature abnormality warning threshold T v .

[0150] Using the overload warning threshold T o It can monitor and trigger warnings in real time, and specifically monitor the power load continuously in real time, and compare the power load value monitored in real time with the overload warning threshold. Once the real-time power load reaches or exceeds the overload warning threshold, the overload warning signal is immediately triggered. This allows power system operation and maintenance personnel to know in time that the power supply is about to face overload risks, and make preparations in advance, such as adjusting power generation plans, reasonably allocating power resources, etc., to avoid serious consequences such as damage to power equipment and power outages due to overload, and ensure the stable operation of the power system.

[0151] Using the temperature abnormality warning threshold T v It can ensure the safe operation of equipment. Specifically, when the monitored temperature-related data (such as the operating temperature of power equipment, ambient temperature, etc.) reaches or exceeds the abnormal temperature warning threshold, a warning will be issued in time. This enables operation and maintenance personnel to take measures quickly, such as checking whether the equipment cooling system is normal, strengthening ventilation and cooling measures, etc., to prevent the performance of power equipment from being degraded, the life of the equipment from being shortened, or even failures due to excessive temperature, and ensure that the power equipment operates safely and stably in a suitable temperature environment.

[0152] See also Figure 2 , Figure 2The present invention is a flow chart of a power monitoring system based on the influence of ocean thermal anomaly according to an exemplary embodiment, wherein the system comprises:

[0153] The data acquisition module 20 is used to collect multi-source data using a variety of sensors; wherein the multi-source data includes: ocean temperature data, ocean current data, atmospheric temperature data, and atmospheric pressure data, voltage, current, power data of each substation, and load data of the power line;

[0154] The ocean thermal anomaly analysis module 21 is used to construct a composite sequence using the ocean temperature data and the ocean current data, perform spectrum decomposition on the composite sequence to obtain all detail components and approximate components, perform feature reconstruction using all detail components and approximate components, and obtain ocean thermal anomaly feature indicators using the result of the feature reconstruction, preset singular value decomposition and indicator construction;

[0155] The power load prediction module 22 is used to construct a basic prediction model input vector by using the ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, and to correct the basic prediction model input vector according to the ocean thermal anomaly state to obtain a corrected input vector;

[0156] Using the corrected input vector and the prediction model based on the long short-term memory network, a power load prediction model is constructed;

[0157] Using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data of the period of ocean thermal anomaly, the power load prediction model is trained using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load;

[0158] The monitoring and early warning module 23 is used to use the prediction results of the power load and the ocean thermal anomaly characteristic index to set early warning thresholds, including an overload early warning threshold and a temperature anomaly early warning threshold.

[0159] See also Figure 3 , Figure 3 The present invention is a schematic block diagram showing the composition of a power monitoring device based on the influence of ocean thermal anomaly according to an exemplary embodiment, wherein the device comprises:

[0160] A memory 31 on which an executable program is stored;

[0161] The processor 32 is used to execute the executable program in the memory 31 to implement the steps of any one of the above methods.

[0162] The technical solution provided by the embodiments of the present invention may have the following beneficial effects:

[0163] It can be understood that the technical solution provided by the present invention uses multiple sensors to collect multi-source data;

[0164] A composite sequence is constructed by using the ocean temperature data and the ocean current data, and the composite sequence is spectrally decomposed to obtain all detail components and approximate components. All detail components and approximate components are used to perform feature reconstruction, and the result of the feature reconstruction, the preset singular value decomposition and the index construction are used to obtain the characteristic index of ocean thermal anomaly. Afterwards, a basic prediction model input vector is constructed by using the characteristic index of ocean thermal anomaly, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, and the basic prediction model input vector is corrected according to the state of ocean thermal anomaly to obtain the corrected input vector. A power load prediction model is constructed by using the corrected input vector and a prediction model based on a long short-term memory network. The power load prediction model is trained by using the power load data, the atmospheric temperature data, the atmospheric pressure data, and the relevant data of the period of ocean thermal anomaly, and a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain the prediction result of the power load. Finally, the prediction result of the power load and the characteristic index of ocean thermal anomaly are used to set warning thresholds, including overload warning thresholds and temperature anomaly warning thresholds.

[0165] It is understandable that the technical solution provided by the present invention, through the ocean thermal anomaly analysis module, deeply explores the intrinsic connection between ocean temperature and ocean current data, extracts the characteristics of ocean thermal anomalies, and provides a key correction basis for power load forecasting. When ocean thermal anomalies occur, it can timely capture the potential fluctuation trend of power load caused by changes in the ocean environment, avoid prediction deviations caused by ignoring ocean factors, effectively balance power supply and demand, reduce energy waste and power shortages caused by inaccurate load forecasts, and ensure stable operation of the power system. Based on the overload warning threshold and the temperature anomaly warning threshold, it can fully consider the impact of ocean thermal anomalies on various links of the power system, greatly improve the accuracy of the warning, and when the power system faces potential risks, it can timely issue warning signals and provide detailed response strategy suggestions, shorten the troubleshooting and repair time, and improve the operation and maintenance efficiency of the power system.

[0166] Through innovative ocean thermal anomaly analysis algorithms, the characteristics of ocean thermal anomalies can be accurately extracted to provide key correction basis for power load forecasting and improve prediction accuracy. At the same time, a scientific and reasonable power load forecasting model is established to fully consider the combined influence of ocean thermal anomaly factors and the atmospheric environment and the power system's own data. In addition, an early warning threshold based on comprehensive consideration of multiple factors is set to improve the accuracy and timeliness of monitoring and early warning, thereby ensuring the stable operation of the power system under the influence of ocean thermal anomalies and reducing power failures and losses caused by changes in the marine environment.

[0167] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0168] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0169] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0170] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0171] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0172] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0173] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0174] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0175] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A power monitoring method based on the influence of ocean thermal anomalies, characterized in that: The method comprises: Use multiple sensors to collect multi-source data; the multi-source data includes: ocean temperature data, ocean current data, atmospheric temperature data, atmospheric pressure data, voltage, current, power data of each substation, and load data of power lines; Using the ocean temperature data and the ocean current data, a composite sequence is constructed, and the composite sequence is subjected to spectral decomposition to obtain all detail components and approximate components, and all detail components and approximate components are used to perform feature reconstruction, and the result of the feature reconstruction, preset singular value decomposition and index construction are used to obtain an ocean thermal anomaly feature index; Using the ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, a basic prediction model input vector is constructed, and the basic prediction model input vector is corrected according to the ocean thermal anomaly state to obtain a corrected input vector; Using the corrected input vector and the prediction model based on the long short-term memory network, a power load prediction model is constructed; Using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data of the period of ocean thermal anomaly, the power load prediction model is trained using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load; Using the prediction result of the power load and the ocean thermal anomaly characteristic index, warning thresholds are set, including an overload warning threshold and a temperature anomaly warning threshold.

2. The method according to claim 1, characterized in that The method of constructing a composite sequence by using the ocean temperature data and the ocean current data, and performing spectrum decomposition on the composite sequence to obtain all detail components and approximate components includes: S1. Using the ADTF algorithm, define the initial decomposition components: S2. At each decomposition, the instantaneous frequency is calculated based on the Hilbert transform of the current decomposition component; S3. Constructing a time-frequency filtering window according to the instantaneous frequency and the empirical coefficient; S4. Filter the current decomposition component using the time-frequency filter window to obtain detail components and approximate components; S5. Repeat the decomposition steps S2-S4 until the decomposition times reach the upper limit, and obtain all detail components and approximate components.

3. The method according to claim 1, characterized in that The feature reconstruction using all detail components and approximate components includes: Normalize each detail component into a specific form by calculating its mean and standard deviation; Normalize each approximate component into a specific form by calculating its mean and standard deviation; Constructing a feature matrix using the normalized values ​​of the detail components at different time points as row vectors; The normalized approximate component is added as the last row vector of the feature matrix to obtain a new feature matrix.

4. The method according to claim 1, characterized in that: The result of the feature reconstruction, the preset singular value decomposition and the index construction are used to obtain the ocean thermal anomaly feature index, including: Perform singular value decomposition on the new feature matrix to obtain a singular value matrix; Selecting left singular vectors corresponding to a specific number of singular values ​​in the singular value matrix for feature reconstruction; Elements are selected from the reconstructed eigenvector and combined with the corresponding weight coefficients to construct the ocean thermal anomaly characteristic index.

5. The method according to claim 1, characterized in that The method uses the ocean thermal anomaly state, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data to construct a basic prediction model input vector, and corrects the basic prediction model input vector according to the ocean thermal anomaly characteristic index to obtain a corrected input vector, including: The ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data are obtained, and the lag order is determined according to the autocorrelation analysis of the historical power load data to construct the basic prediction model input vector, as shown in the following formula: X(t)=[L(t-1),L(t-2),…,L(tn),A(t),P(t)] (1) Where n is the lag order determined based on the autocorrelation analysis of historical power load data, A(t) represents the atmospheric temperature data, P(t) represents the atmospheric pressure data, and L(t) represents the power system operation data; According to the ocean thermal anomaly state, when the preset conditions are met, the ocean thermal anomaly correction term is introduced into the basic prediction model input vector to obtain the corrected input vector, as shown in the following formula: X m (t)=[L(t-1),L(t-2),…,L(t-n),A(t),P(t),γF(t)] (2) in, I represents the characteristic index of ocean thermal anomaly, and γ is the weight coefficient of the impact of ocean thermal anomaly on power load.

6. The method according to claim 1, characterized in that The method of using the corrected input vector and the prediction model based on the long short-term memory network to construct a power load prediction model includes: Built on LSTM network, including input layer, hidden layer and output layer; The number of neurons in the input layer is based on the corrected input vector X m The dimension of (t) is determined by X m The number of elements in (t) is the same; The hidden layer is set to a multi-layer structure and the number of layers is determined by the mean square error as the evaluation index. Each LSTM unit in the hidden layer contains an input gate i t , Forget gate t , output gate o t and memory unit c t , hidden state h t , and calculated according to the following formula: where x t The input vector X m (t) value at time t, h t-1 is the hidden state at the previous moment, W i ,W f ,W o ,W c are the weight matrices corresponding to the gate and memory unit, respectively, i ,b f ,b o ,b c is the bias vector, ⊙ represents element-by-element multiplication, σ is the sigmoid function, and the number of neurons in the output layer is 1, corresponding to the predicted power load value.

7. The method according to claim 1, characterized in that The power load prediction model is trained by using the power load data, the atmospheric temperature data, the atmospheric pressure data and the relevant data of the ocean thermal anomaly period, using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load, including: The power load prediction model is trained using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data during the period of ocean thermal anomaly, and the connection weights of the neural network are optimized using an adaptive moment estimation optimization algorithm to update network parameters; Using the preset total number of training steps and the current number of training steps, the learning rate is adjusted through a dynamic learning rate adjustment strategy to obtain the prediction result of the power load; The power load prediction model is trained by using the power load data, the atmospheric temperature data, the atmospheric pressure data and the relevant data of the ocean thermal anomaly period, and the connection weights of the neural network are optimized by using the adaptive moment estimation optimization algorithm to update the network parameters, including: Initialize the first-order moment vector and the second-order moment vector to 0, and set the learning rate to the initial value η0; The gradient is calculated according to the preset loss function; the loss function is as follows: Among them, y i is the true value, is the predicted value, θ is the network parameter; The first-order moment vector is updated using a preset first formula; wherein the preset first formula is as follows: m t =β1m t-1 +(1-β1)g t (5) Where β1 is a hyperparameter used to control the moving average weight of the gradient; The second-order moment vector is updated using a preset second formula; wherein the preset second formula is as follows: Where β2 is a hyperparameter used to control the moving average weight of the squared gradient; respectively performing deviation correction on the first-order moment vector and the second-order moment vector; The network parameters are updated according to the first-order moment vector and the second-order moment vector after the deviation correction.

8. The method according to claim 6, characterized in that The method uses the preset total number of training steps and the current number of training steps, and adjusts the learning rate through a dynamic learning rate adjustment strategy to obtain a prediction result of the power load, including: Using the preset total number of training steps and the current number of training steps, as shown in the following formula, the learning rate is adjusted through the dynamic learning rate adjustment strategy to obtain the prediction result of the power load; In the early stage of training When the learning rate η t =η0; In the middle of training hour, In the later stages of training hour, Where N is the total number of training steps, t is the current number of training steps, η 0为 Initial learning rate.

9. A power monitoring system based on the influence of ocean thermal anomalies, applied to the power monitoring method based on the influence of ocean thermal anomalies as claimed in any one of claims 1 to 8, characterized in that: The system comprises: The data acquisition module is used to collect multi-source data using a variety of sensors; the multi-source data includes: ocean temperature data, ocean current data, atmospheric temperature data, atmospheric pressure data, voltage, current, power data of each substation, and load data of the power line; The ocean thermal anomaly analysis module is used to construct a composite sequence using the ocean temperature data and the ocean current data, perform spectrum decomposition on the composite sequence to obtain all detail components and approximate components, perform feature reconstruction using all detail components and approximate components, and obtain ocean thermal anomaly feature indicators using the result of the feature reconstruction, preset singular value decomposition and indicator construction; The power load forecasting module is used to construct a basic forecasting model input vector by using the ocean thermal anomaly characteristic index, the atmospheric temperature data, the atmospheric pressure data, and the preset power system operation data, and to correct the basic forecasting model input vector according to the ocean thermal anomaly state to obtain a corrected input vector; Using the corrected input vector and the prediction model based on the long short-term memory network, a power load prediction model is constructed; Using the power load data, the atmospheric temperature data, the atmospheric pressure data, and relevant data of the period of ocean thermal anomaly, the power load prediction model is trained using a preset adaptive moment estimation optimization algorithm and a dynamic learning rate adjustment strategy to obtain a prediction result of the power load; The monitoring and early warning module is used to use the prediction results of the power load and the ocean thermal anomaly characteristic indicators to set early warning thresholds, including overload early warning thresholds and temperature anomaly early warning thresholds.

10. Power monitoring equipment based on the influence of ocean thermal anomalies, characterized in that: The device comprises: a memory having an executable program stored therein; A processor, configured to execute the executable program in the memory to implement the steps of the method according to any one of claims 1 to 8.

Images