Industrial important security load safe operation potential assessment method and related equipment

Abstract

The invention discloses an industrial important security load safe operation potential assessment method and related equipment, and relates to the field of load monitoring. Comprising the steps that industrial total load data is collected in a non-intrusive mode, a security load super-state set is established based on improved K-means clustering and a finite state machine, a hidden Markov model is introduced to achieve key security load power decomposition, short-term power prediction is conducted on key security loads through a long-short-term memory network, and the security load super-state set is obtained. On the basis of considering equipment rated power, voltage allowable deviation and emergency starting time requirements, safety margin indexes such as a power margin, a response time margin and a voltage stability margin are calculated, and the safety margins before and after regulation and control are compared to obtain the safety operation potential. According to the invention, fine identification and prospective safety evaluation of the industrial security load are realized without adding an equipment side monitoring channel.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of load monitoring, in particular to an industrial important security load safe operation potential evaluation method and related equipment. BACKGROUND

[0002] Industrial important security load (such as production line key equipment, emergency power supply system, cooling circulating device, safety monitoring system, etc.) is the core load to protect the continuity of industrial production, personnel safety and major equipment protection, and its operation state is directly related to the safety and stability of the industrial system. Current evaluation of industrial security load relies on single device monitoring data, without considering the correlation between loads and the influence of power grid fluctuations on overall safety, making it difficult to accurately evaluate the safety operation margin of the load under power grid disturbance. SUMMARY

[0003] The technical problem to be solved by the present application is that the current industrial security load regulation potential does not consider the correlation between loads and the influence of power grid fluctuations on overall safety. The purpose is to provide an industrial important security load safe operation potential evaluation method and related equipment, which solves the problem of how to accurately evaluate the safety operation margin of industrial security load regulation under power grid disturbance.

[0004] The present application is realized by the following technical scheme:

[0005] In a first aspect, the present application provides an industrial important security load safe operation potential evaluation method, comprising:

[0006] By non-invasively collecting total load data of the industrial system, based on the state clustering results of the historical power data of multiple security load devices, the operating states of each security load device are coded and a finite state machine is established, and the load super state set reflecting the cooperative operation condition of multiple devices is constructed according to the combination of the state codes of each security load device;

[0007] A hidden Markov model is established with the load super state set as the hidden state set and the industrial total load power sequence as the observation sequence, and according to the preset initial probability distribution, state transition probability and observation probability, the optimal super state sequence is solved, and the power time sequence of at least one key security load is obtained from the industrial total load sequence by combining the power representative value corresponding to the different states of each security load device;

[0008] A long short-term memory network model is constructed, the industrial total load power sequence and the power time sequence of the key security load are taken as input features, and the predicted power value of the key security load at each time in the target prediction period is output;

[0009] According to the preset industry safety standard and the equipment operation requirement, an industrial security load safe operation model is constructed, based on the predicted power of the key security load and the safety threshold such as the equipment rated power, the voltage allowable deviation and the emergency starting time requirement, the safety margin index representing the power margin, the response time margin and the voltage stability margin is calculated, and the safety margin before and after regulation is compared to obtain the safety operation potential of the key security load.

[0010] Further, the construction of the load super state comprises:

[0011] For the historical power time sequence of each type of security load equipment, under the constraints of the equipment rated power and the preset industry safety threshold, an improved K-means clustering method is adopted, taking the typical operation power, standby power and starting peak power of the equipment as the initial clustering center, and the weighted Euclidean distance is used to measure the distance between different power data points and each clustering center to obtain the power representative value of each operating state.

[0012] The state codes of various security loads in the industrial scene are combined to form a vector composed of the state codes of various security load equipment, and the vector is defined as the load super state, which is used to represent the cooperative operation condition of multiple security load equipment.

[0013] Further, the key security load is decomposed based on the hidden Markov model, comprising:

[0014] The set of load super states is defined as the hidden state set, and the industrial total load power sequence is defined as the observation set, and the initial probability distribution, the state transition probability matrix and the observation probability matrix of observing different total power under each hidden state are obtained based on the historical operation data statistics;

[0015] The Viterbi algorithm is used to solve the optimal hidden state sequence corresponding to the observation set, and the decomposition power of the key security load at each time is calculated according to the state of each security load equipment and the power representative value in the optimal hidden state sequence.

[0016] Further, the long short-term memory network model comprises:

[0017] The input layer takes the industrial total load power and the key security load power decomposed by the hidden Markov model as the input features;

[0018] The regularization layer is used to extract the periodic fluctuation characteristics and mutation characteristics in the load time sequence to suppress overfitting;

[0019] The fully connected output layer is used to map the long short-term memory network layer output to the predicted power;

[0020] The long short-term memory network model is trained offline based on the historical operation data containing normal conditions and fault conditions.

[0021] Further, during the training and prediction of the long short-term memory network model, production plan information representing industrial production shift scheduling and maintenance plans and voltage stability indexes representing power grid operating conditions are further taken as auxiliary input features to represent the influence of production plans on load power changes and the influence of power grid conditions on the operating state of security loads.

[0022] Further, the calculation of the safety margin index in the safe operation model includes:

[0023] obtaining a power margin based on the difference between the rated power of the equipment and the actual output power or the predicted output power;

[0024] obtaining a response time margin based on the difference between the emergency start requirement time and the corresponding actual response time; and

[0025] obtaining a voltage stability margin based on the voltage deviation between the actual output voltage and the grid voltage reference value and comparing it with the preset maximum allowed voltage deviation.

[0026] Further, the comparison of the safety margins before and after the regulation to obtain the safety operation potential includes:

[0027] determining the safety margin before regulation based on the minimum value among the power margin, the response time margin and the voltage stability margin when no regulation measures are implemented;

[0028] recomputing the power margin, the response time margin and the voltage stability margin under the operating conditions after adjusting the load rate of the key security load equipment and / or starting the standby security power supply based on the short-term power prediction result, and determining the safety margin after regulation based on the minimum value among the three;

[0029] taking the difference between the safety margin after regulation and the safety margin before regulation as the safety operation potential of the key security load.

[0030] In a second aspect, the present application also provides an industrial important security load safety operation potential evaluation system for implementing the industrial important security load safety operation potential evaluation method as described above, comprising:

[0031] an industrial total load data acquisition module for acquiring total load power time series data of an industrial power system through a non-intrusive measurement method;

[0032] The load pretreatment and superstate construction module is used for state clustering based on historical power data of various security load devices, obtaining power representative values of various operating states, encoding operating states of various security load devices and establishing a finite state machine, and constructing a load superstate set representing a multi-device collaborative operation condition according to combinations of state codes of various security load devices.

[0033] The hidden Markov model load decomposition module is used for establishing a hidden Markov model with the load superstate set as a hidden state set and with an industrial total load power time sequence as an observation sequence, solving an optimal hidden state sequence based on a preset initial probability distribution, a state transition probability and an observation probability, and decomposing at least one power time sequence of a key security load from the industrial total load sequence according to states of various security load devices in the optimal hidden state sequence and power representative values thereof.

[0034] The long short-term memory network prediction module is used for constructing a long short-term memory network model, taking the industrial total load power time sequence, the power time sequence of the key security load and optional production plan information and power grid operation indexes as input features, and outputting predicted power values of the key security load at various time points in a target prediction period.

[0035] The safe operation potential evaluation module is used for constructing an industrial security load safe operation model according to preset industry safety standards and device operation requirements, calculating safety margin indexes representing power margin, response time margin and voltage stability margin based on predicted power of the key security load and a safety threshold, and determining a safe operation potential of the key security load by comparing safety margins before and after regulation and control.

[0036] In a third aspect, the present application further provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to realize the industrial important security load safe operation potential evaluation method as described above.

[0037] In a fourth aspect, the present application further provides a computer readable storage medium, which stores a computer program, wherein the program is executable on a processor to realize the industrial important security load safe operation potential evaluation method as described above.

[0038] Compared with the prior art, the present application has the following advantages and beneficial effects:

[0039] The application only collects total load power data of a total electric meter or a total bus of a plant, and completes key security load power decomposition by using a finite state machine, load super state modeling and a hidden Markov model without increasing equipment-side collection channels and without modifying existing security equipment, compared with a scheme relying on single equipment monitoring, greatly reduces implementation cost and modification difficulty, and improves load identification coverage and accuracy.

[0040] The application includes scenes such as main power failure, emergency power start, and cooling system linkage into a unified state space by constructing a load super state and state transition probability of multiple equipment cooperative operation, and solves an optimal super state sequence by using a Viterbi algorithm, thereby realizing time sequence decomposition of a single key security load power under the condition of reserving linkage logic between equipment, and is more suitable for an industrial scene with strong coupling of multiple security equipment compared with load analysis based only on single equipment characteristics.

[0041] By taking total load power and key security load power obtained by decomposition as main features, combining production shift arrangement, maintenance plan and power grid voltage stability index, a long short-term memory network is constructed and trained to predict security load power in a future short period, so that security margin evaluation is moved from post-event analysis to pre-event prediction, and has better adaptability to extreme disturbance scenes under training of historical data containing fault conditions.

[0042] The application constructs an industrial security load dynamic operation equation on the basis of referring to an equivalent thermal parameter model, defines three types of quantitative indexes of power margin, response time margin and voltage stability margin in combination with industry safety standards, and obtains security operation potential by comparing changes in security margin before and after regulation and control, so that the evaluation result can directly reflect the improvement degree of safety level of measures such as load regulation and standby power source start. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the example embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be considered as a limitation to the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings. In the drawings:

[0044] Figure 1 The step flow chart of the embodiment 1 of the application. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the application more clear and obvious, the following will further describe the application in combination with embodiments and drawings. The illustrative embodiments of the application and the description thereof are only used to explain the application, and should not be considered as a limitation to the application.

[0046] Example 1

[0047] A method for assessing the safe operating potential of critical industrial security loads, such as Figure 1 As shown, it includes:

[0048] Industrial total load data preprocessing and load overstate construction steps: collect total load data of industrial system in a non-intrusive manner, based on the state clustering results of historical power data of multiple security load devices, encode the operating state of each security load device and establish a finite state machine, and construct a load overstate set reflecting the collaborative operation of multiple devices according to the combination of state codes of each security load device.

[0049] Key security load decomposition steps based on hidden Markov model: Using the load overstate set as the hidden state set and the industrial total load power sequence as the observation sequence, establish a hidden Markov model, solve the optimal overstate sequence according to the preset initial probability distribution, state transition probability and observation probability, and combine the power representative values ​​corresponding to different states of each security load equipment to decompose at least one key security load power time series from the industrial total load sequence.

[0050] The steps for short-term power prediction based on long short-term memory networks are as follows: Construct a long short-term memory network model, take the power sequence of the total industrial load and the power time sequence of the critical security load as input features, and output the predicted power value of the critical security load at each time during the target prediction period.

[0051] The steps for assessing the safe operating potential based on safety thresholds are as follows: An industrial safety load safety operation model is constructed based on preset industry safety standards and equipment operation requirements. Based on the predicted power of the key safety load and safety thresholds such as equipment rated power, voltage allowable deviation, and emergency start-up time requirements, safety margin indices characterizing power margin, response time margin, and voltage stability margin are calculated. The safety margins before and after regulation are compared to obtain the safe operating potential of the key safety load.

[0052] This invention collects total load power data from the plant's main electricity meter or key busbars. Without increasing the number of data collection channels on the equipment side or modifying existing security equipment, it uses finite state machines, load over-state modeling, and hidden Markov models to decompose the power of key security loads. Compared with solutions that rely on single-equipment monitoring, this invention significantly reduces implementation costs and modification difficulty, while improving the coverage and accuracy of load identification.

[0053] In this embodiment, the construction of the load over-state includes:

[0054] For the historical power time series of each type of security load equipment, under the constraints of the equipment rated power and the preset industry safety threshold, an improved K-means clustering method is adopted with the typical operating power, standby power and peak starting power of the equipment as the initial cluster centers. The weighted Euclidean distance is used to measure the distance between different power data points and each cluster center to obtain the power representative value of each operating state.

[0055] The state codes of various security loads in industrial scenarios are combined to form a vector composed of the state codes of each security load device. The vector is defined as the load over-state, which is used to characterize the collaborative operation of multiple security load devices.

[0056] Specifically, industrial safety load equipment has multi-state characteristics (such as "operating / standby / fault / maintenance", "full load / half load / no load", etc.). An improved K-means algorithm is used to cluster the historical power data of a single device.

[0057] Input: Historical power sequence (length T) of a single emergency load device (such as an emergency generator or a critical pump set);

[0058] Clustering logic: Constrained by the rated power of the equipment and industry safety thresholds, the initial cluster centers are combined with prior knowledge (rather than randomly selected), such as the "start-up peak power", "stable operating power" and "standby power" of emergency generators as the initial centers;

[0059] Distance calculation: Weighted Euclidean distance is used (the weights reflect the degree of impact of different states on safety, such as "fault state" having a higher weight than "standby state"): (In the formula, For state weights, For power data points, (the center of the k-th cluster);

[0060] Convergence condition: When the SSE (sum of squared errors) converges or reaches the iteration limit, the cluster center is the power representative value of each state of the device (e.g., emergency generator "stable operating power = 200kW" "standby power = 5kW").

[0061] Industrial safety load devices exhibit strong interdependencies (e.g., a cooling system failure could cause production line shutdowns), necessitating the construction of a "super-state" that reflects the collaboration of multiple devices.

[0062] Suppose there are N types of security load devices in an industrial scenario (such as emergency power supplies, cooling pumps, and security monitoring systems). The nth type of load has If there are 10 states, then the superstate is represented as: , (In the formula, (This is the status code for the nth type of load, where 0 indicates off, 1 indicates running, 2 indicates fault, etc.)

[0063] For example: superstate This indicates that "emergency power supply is running, cooling pump is running, and safety monitoring system is shut down."

[0064] This invention constructs a load overstate and its state transition probability for multi-device collaborative operation, incorporating scenarios such as main power failure, emergency power start-up, and cooling system linkage into a unified state space. It then uses the Viterbi algorithm to solve for the optimal overstate sequence, thereby achieving time-series decomposition of the power of a single critical security load while preserving the linkage logic between devices. Compared to load analysis based solely on single-device characteristics, this invention is more suitable for industrial scenarios with strong coupling of multiple security devices.

[0065] In this embodiment, the decomposition of critical security loads based on a hidden Markov model includes:

[0066] The set of load over-states is defined as the hidden state set, and the industrial total load power sequence is defined as the observation set. The initial probability distribution, the state transition probability matrix, and the observation probability matrix of different total power observed in each hidden state are obtained based on historical operating data.

[0067] The Viterbi algorithm is used to solve for the optimal hidden state sequence corresponding to the observation set, and the decomposed power of the critical security load at each time point is calculated based on the state of each security load device and its power representative value in the optimal hidden state sequence.

[0068] HMM (Hidden Markov Model) Parameter Definition

[0069] State set: Superstate (J represents the total number of superstates);

[0070] Observation set: Total load power sequence ;

[0071] Initial probability distribution : Reflects the probability of each super-state under the initial state of industrial production (e.g., the probability of the super-state is higher when "all equipment is running" at startup): , For super state (where V is the number of samples and V is the total number of samples).

[0072] State transition probability A: Reflects the linkage characteristics of industrial loads (e.g., the transition probability of main power failure → emergency power activation): , For super state Transferred to (frequency)

[0073] Observation probability B: Overstate The total power observed below The probability of: ;

[0074] The Viterbi algorithm is used to solve the optimal hyperstate sequence. Combined with the representative value of the state power of a single device, the power of each security load at each moment is deduced (e.g., the emergency power of 200kW and the cooling pump power of 50kW are decomposed from the hyperstate \(S=(1,1,0)\).

[0075] In this embodiment, the Long Short-Term Memory (LSTM) network model includes:

[0076] The input layer takes the total industrial load power and the critical security load power obtained by the decomposition of the Hidden Markov Model as input features;

[0077] Regularization layers are used to extract periodic fluctuation and abrupt change features from the load time series to suppress overfitting;

[0078] Fully connected output layer: used to map the output of the Long Short Time Memory network layer to the predicted power;

[0079] The Long Short-Term Memory (LSTM) network model is trained offline based on historical operating data that includes both normal and fault conditions.

[0080] When training and predicting the Long Short-Term Memory Network model, production plan information representing industrial production shifts and maintenance plans, as well as voltage stability indicators representing power grid operating conditions, are further used as auxiliary input features to characterize the impact of production plans on load power changes and the impact of power grid operating conditions on the operating status of security loads.

[0081] LSTM model structure design, including:

[0082] Input layer: 2 types of features - total industrial load power and critical security load power (such as emergency power supply power);

[0083] LSTM layer: 1-layer LSTM network (128 neurons) to capture the periodicity of the load (such as load fluctuations caused by production shifts) and the suddenness (such as equipment start-up and shutdown).

[0084] Regularization: Add a Dropout layer (dropout rate of 0.3, higher than the original patent, due to more severe fluctuations in industrial loads);

[0085] Output layer: Fully connected layer + regression output, predicts the power of the security load at future moments (such as the power of the emergency power supply in the next 15 minutes).

[0086] Prediction optimization includes:

[0087] Input feature enhancement: Add industrial production plans (such as "downtime for maintenance") and power grid voltage stability indicators to improve prediction accuracy;

[0088] Model training: The model is trained using historical industrial load data (including fault conditions) to adapt to extreme scenarios.

[0089] By using total load power and key safety load power obtained from decomposition as the main features, combined with production shift arrangements, maintenance plans and power grid voltage stability indicators, a long short-term memory network is constructed and trained to predict safety load power in the short term. This shifts safety margin assessment from post-event analysis to pre-event prediction, and with training on historical data including fault conditions, it has better adaptability to extreme disturbance scenarios.

[0090] In this embodiment, the calculation of the safety margin index in the safe operation model includes:

[0091] The power margin is obtained based on the difference between the rated power of the equipment and the actual or predicted output power.

[0092] The response time margin is derived from the difference between the required emergency activation time and the corresponding actual response time; and

[0093] The voltage stability margin is obtained by comparing the voltage deviation between the actual output voltage and the grid voltage reference value with the preset maximum allowable voltage deviation.

[0094] Constructing the dynamic operating equations for security loads (taking emergency power supply as an example):

[0095]

[0096] In the formula: This is an indicator of the stability of the output voltage of the emergency power supply (reflecting the safety status). Where C is the grid voltage reference value, and C is the equivalent heat capacity for equipment voltage regulation. P represents the power conversion efficiency, P represents the power output power, and s represents the operating state (1 = running, 0 = off).

[0097] In this embodiment, comparing the safety margin before and after adjustment to obtain the safe operating potential includes:

[0098] When no control measures are implemented, the safety margin before control is determined based on the minimum value among the power margin, response time margin, and voltage stability margin.

[0099] Under the operating conditions after adjusting the load rate of critical security load equipment and / or starting the backup security power supply based on the short-term power prediction results, the power margin, response time margin and voltage stability margin are recalculated, and the safety margin after regulation is determined based on the minimum value of the three.

[0100] The difference between the safety margin after regulation and the safety margin before regulation is taken as the safe operating potential of the critical security load.

[0101] Based on industry safety standards, define safety margin indicators:

[0102] Power margin: , ( Rated power of the equipment (The actual power must be ≥0).

[0103] Response time margin: , ( To meet the time requirements for emergency activation, (The actual response time must be ≥0).

[0104] Voltage stability margin: ( (To allow the maximum voltage deviation).

[0105] Assess potential by comparing the safety margins before and after regulation:

[0106] Safety margin before regulation: ;

[0107] Safety margin after adjustment (e.g., adjusting equipment load rate, starting backup power): ;

[0108] Safe operation potential: (A positive value indicates that regulation can increase the safety margin).

[0109] This invention constructs a dynamic operating equation for industrial safety load based on a reference equivalent thermal parameter model. It combines industry safety standards to define three types of quantitative indicators: power margin, response time margin, and voltage stability margin. By comparing the changes in safety margin before and after regulation, it obtains the safe operating potential, enabling the evaluation results to directly reflect the degree of improvement in safety level caused by measures such as load adjustment and starting backup power.

[0110] Example 2

[0111] A system for assessing the safe operation potential of critical industrial safety loads includes a load preprocessing and over-state construction module, a hidden Markov model load decomposition module, a long short-time memory network prediction module, and a safe operation potential assessment module, wherein:

[0112] The industrial total load data acquisition module is used to acquire time series data of total load power of industrial power systems through non-intrusive measurement methods.

[0113] The load preprocessing and overstate construction module is used to perform state clustering based on the historical power data of various security load devices, obtain the power representative value of each operating state, encode the operating state of each security load device and establish a finite state machine, and construct a load overstate set representing the collaborative operation of multiple devices according to the combination of the state codes of each security load device.

[0114] The Hidden Markov Model Load Decomposition Module is used to establish a Hidden Markov Model with the load overstate set as the hidden state set and the industrial total load power time series as the observation sequence. Based on the preset initial probability distribution, state transition probability and observation probability, it solves the optimal hidden state sequence and decomposes the power time series of at least one key security load from the industrial total load sequence according to the state of each security load device and its power representative value in the optimal hidden state sequence.

[0115] The Long Short-Term Memory Network Prediction Module is used to construct a Long Short-Term Memory Network Model. It takes the power time series of the total industrial load, the power time series of the key security load, and optional production plan information and power grid operation indicators as input features, and outputs the predicted power values ​​of the key security load at each time point within the target prediction period.

[0116] The safe operation potential assessment module is used to construct an industrial safety load safe operation model based on preset industry safety standards and equipment operation requirements. Based on the predicted power and safety threshold of the key safety load, it calculates safety margin indicators that characterize power margin, response time margin, and voltage stability margin, and determines the safe operation potential of the key safety load by comparing the safety margins before and after regulation.

[0117] In this embodiment, the load preprocessing and over-state construction module is specifically used to: for the historical power time series of each type of security load equipment, under the constraints of the equipment's rated power and preset industry safety threshold, adopt an improved K-means clustering method with the typical operating power, standby power and startup peak power of the equipment as the initial cluster centers, use weighted Euclidean distance to measure the distance between different power data points and each cluster center to obtain the power representative value of each operating state, and combine the state codes of various types of security loads to form a load over-state, which is used to characterize the collaborative operating conditions of multiple security load equipment.

[0118] In this embodiment, the hidden Markov model load decomposition module is specifically used to: define the load overstate set as the hidden state set, define the industrial total load power sequence as the observation set, obtain the initial probability distribution, state transition probability matrix and observation probability matrix based on historical operating data, solve the optimal hidden state sequence using the Viterbi algorithm, and calculate the decomposed power of the key security load at each moment based on the state of each security load device and its power representative value in the optimal hidden state sequence.

[0119] In this embodiment, the Long Short-Term Memory (LSTM) network prediction module includes an input layer, an LSM network layer, a regularization layer, and a fully connected output layer. The input layer uses the total industrial load power and the critical safety load power as input features. The regularization layer is used to extract periodic fluctuation features and abrupt change features in the load time series to suppress overfitting. The fully connected output layer is used to map the output of the LSM network layer to the predicted power. The LSM network prediction module is trained offline based on historical operating data including normal operating conditions and fault operating conditions.

[0120] In this embodiment, the Long Short-Term Memory Network prediction module is further configured to use production plan information and voltage stability index as auxiliary input features during the training and prediction process. The production plan information includes industrial production shift arrangements and maintenance plans, and the voltage stability index is used to characterize the power grid operating conditions.

[0121] In this embodiment, the safe operation potential assessment module is specifically used for: calculating the power margin based on the difference between the rated power of the equipment and the actual output power or the predicted output power; calculating the response time margin based on the difference between the emergency start-up required time and the corresponding actual response time; calculating the voltage stability margin based on the voltage deviation between the actual output voltage and the grid voltage reference value and comparing it with the preset maximum allowable voltage deviation; determining the safety margin before regulation based on the minimum value among the three when no regulation measures are implemented; and recalculating the safety margin and determining the safety margin after regulation after adjusting the load rate of the critical security load equipment and / or starting the backup security power supply according to the short-term power prediction results, and taking the difference between the two as the safe operation potential of the critical security load.

[0122] Example 3

[0123] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for assessing the safe operating potential of industrial critical security loads as described in Example 1.

[0124] Example 4

[0125] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the method for assessing the safe operating potential of industrial critical security loads as described in Example 1.

[0126] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0127] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0128] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0129] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0130] Those skilled in the art will understand that all or part of the steps in the above facts and methods can be implemented by a program instructing related hardware. The program or the program described therein can be stored in a computer-readable storage medium. When the program is executed, it includes the following steps: at this time, the corresponding method steps are introduced. The storage medium can be ROM / RAM, magnetic disk, optical disk, etc.

[0131] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating the safe operation potential of an important security load in an industry, characterized in that, The method comprises the following steps: encoding the operating states of each security load device based on the state clustering results of the historical power data of the plurality of security load devices, and establishing a finite state machine, and constructing a load super state set reflecting the operating conditions of the plurality of security load devices according to the combination of the state codes of the security load devices; establishing a hidden Markov model with the load super state set as the hidden state set and the industrial total load power sequence as the observation sequence, solving the optimal super state sequence according to the preset initial probability distribution, state transition probability and observation probability, and decomposing at least one key security load power time sequence from the industrial total load sequence in combination with the power representative values corresponding to the different states of each security load device; constructing a long short-term memory network model, taking the industrial total load power and the key security load power decomposed by the hidden Markov model as input features, and outputting the predicted power values of the key security load at each time in the target prediction period; constructing an industrial security load safe operation model according to the preset industry safety standards and equipment operation requirements, calculating the safety margin indicators representing the power margin, response time margin and voltage stability margin based on the predicted power of the key security load and the safety threshold, and comparing the safety margins before and after regulation to obtain the safety operation potential of the key security load, wherein the safety threshold comprises a device rated power threshold, a voltage allowable deviation threshold and an emergency start time requirement threshold.

2. The method for assessing the safe operating potential of an industrial critical security load according to claim 1, characterized in that The construction of the load super state comprises: under the constraints of the device rated power and the preset industry safety threshold, the power representative values of each operating state are obtained by using an improved K-means clustering method with the device typical operating power, standby power and starting peak power as the initial clustering centers, and the weighted Euclidean distance is used to measure the distance between different power data points and each clustering center; combining the state codes of each type of security load in the industrial scene to form a vector composed of the state codes of each security load device, and defining the vector as a load super state for representing the coordinated operation conditions of the plurality of security load devices.

3. The method for evaluating the safe operation potential of an industrial critical security load according to claim 1, characterized in that, The key security load decomposition based on the hidden Markov model comprises: defining the load super state set as the hidden state set and the industrial total load power sequence as the observation set, obtaining the initial probability distribution, the state transition probability matrix and the observation probability matrix of different total powers under each hidden state based on the historical operation data statistics; solving the optimal hidden state sequence corresponding to the observation set by using the Viterbi algorithm, and calculating the decomposed power of the key security load at each time according to the states of each security load device in the optimal hidden state sequence and the power representative values thereof.

4. The method for assessing the safe operating potential of an industrial critical security load according to claim 1, characterized in that The long short-term memory network model comprises: an input layer taking the industrial total load power and the key security load power decomposed by the hidden Markov model as input features; a regularization layer for extracting periodic fluctuation features and mutation features in the load time sequence to suppress overfitting; a fully connected output layer for mapping the long short-term memory network layer output to a predicted power. The long short-term memory network model is trained offline based on historical operation data, and the historical operation data includes normal working conditions and fault working conditions.

5. The method for evaluating the safe operating potential of an industrial critical security load according to claim 1, characterized in that, During the training and prediction of the long short-term memory network model, production plan information and a voltage stability index are further used as auxiliary input features, the production plan information includes industrial production shift arrangement and maintenance plan, and the voltage stability index is used to represent the power grid operation condition.

6. The method for assessing the safe operating potential of an industrial critical security load according to claim 1, characterized in that The calculation of the safety margin index in the safe operation model includes: a power margin is obtained based on the difference between the rated power of the equipment and the actual output power or the predicted output power; a response time margin is obtained based on the difference between the emergency start requirement time and the corresponding actual response time; a voltage stability margin is obtained by comparing the voltage deviation between the actual output voltage and the power grid voltage reference value with the preset maximum allowed voltage deviation.

7. The method for assessing the safe operating potential of an industrial critical security load according to claim 1, characterized in that The comparison of the safety margins before and after the regulation to obtain the safe operation potential includes: when no regulation measures are implemented, the pre-regulation safety margin is determined based on the minimum value among the power margin, the response time margin and the voltage stability margin; after the load rate of the key security load equipment and / or the standby security power supply are adjusted based on the short-term power prediction result, the power margin, the response time margin and the voltage stability margin are recalculated, and the post-regulation safety margin is determined based on the minimum value among the three; the difference between the post-regulation safety margin and the pre-regulation safety margin is taken as the safe operation potential of the key security load.

8. An industrial important security load safe operation potential evaluation system, characterized in that, It includes: a load preprocessing and super state construction module, which is used to encode the operating state of each security load equipment and establish a finite state machine based on the state clustering results of the historical power data of a plurality of security load equipment, and construct a load super state set reflecting the coordinated operation condition of multiple equipment according to the combination of the state codes of each security load equipment; a hidden Markov model load decomposition module, which is used to establish a hidden Markov model with the load super state set as the hidden state set and the industrial total load power sequence as the observation sequence, solve the optimal super state sequence according to the preset initial probability distribution, state transition probability and observation probability, and decompose at least one power time sequence of the key security load from the industrial total load sequence in combination with the power representative value corresponding to the different states of each security load equipment; a long short-term memory network prediction module, which is used to construct a long short-term memory network model, take the industrial total load power sequence and the power time sequence of the key security load as input features, and output the predicted power value of the key security load at each time in the target prediction period; a safe operation potential evaluation module, which is used to construct an industrial security load safe operation model according to the preset industry safety standard and equipment operation requirement, calculate the safety margin index representing the power margin, the response time margin and the voltage stability margin based on the predicted power of the key security load and the safety threshold, and determine the safe operation potential of the key security load by comparing the safety margins before and after the regulation.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor implements the industrial important security load safe operation potential evaluation method according to any one of claims 1 to 7 when executing the program.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by the processor, implements the industrial important security load safe operation potential evaluation method according to any one of claims 1 to 7.

Images