A method and system for heavy-load query control of a ship power management system

Abstract

This invention relates to the field of load control technology, specifically to a heavy-load interrogation control method and system for a ship power management system. In this invention, a Monte Carlo algorithm is introduced to construct a state transition structure. State behavior samples within a capacity range are defined, and transition paths under combinations of actions and capacity are sampled. The capacity consumption response under each action is accumulated, and an action benefit matrix with quantifiable weights is generated through sample averaging. This forms a data-driven load response model, improving the clarity of scheduling action evaluation. After traversing the entire state space, the magnitude of action reward changes is identified, and convergence is determined using a difference threshold. Based on state convergence, a greedy algorithm is called to read the action with the maximum benefit. An action scheduling sequence is constructed sequentially according to state number, improving the symmetry and fit of power allocation, reducing action mismatch rate and allocation deviation rate, and improving power scheduling convergence efficiency.

Description

Technical Field

[0001] This invention relates to the field of load control technology, and in particular to a heavy-load query control method and system for a ship power management system. Background Technology

[0002] The field of load control technology aims to achieve safe, stable and efficient operation of the power system under conditions of limited and fluctuating power supply capacity by scheduling and managing the start-up, shutdown, power level and access sequence of multiple loads. This prevents system overload and equipment damage caused by loads exceeding power supply capacity, optimizes power distribution based on load priority and energy efficiency standards, ensures continuous power supply to critical equipment and important tasks, and maintains dynamic balance of the power system through inquiry, coordination and scheduling mechanisms when sudden heavy load demands occur.

[0003] The purpose of a heavy-load query control method for a ship power management system is to establish a query and judgment mechanism during ship operation to ensure that the power system can reasonably schedule loads and allocate power under limited power supply capacity when high-power loads are connected and load demand surges. This avoids power outages, equipment damage, and system instability caused by instantaneous overloads. By ensuring priority power supply to critical loads and reducing and delaying non-critical loads, the method achieves safety, continuity, and economy of the ship's power system under different operating conditions, thereby improving overall operating efficiency and reliability.

[0004] Existing technologies rely primarily on load priority and energy efficiency standards for regulation, lacking a quantitative mapping of system state and capacity structure. They can only perform start-stop control through fixed rules, making real-time adjustments difficult when load distribution is complex and capacity fluctuations are significant. When high-power loads are concentrated, response lag often occurs, and some systems continue to supply power to non-critical loads, failing to achieve power reduction and peak-shifting access, thus affecting the continuity of power supply for critical tasks. The lack of state transition path records and reward value statistics in the system means that scheduling actions lack a data feedback loop, making dynamic optimization difficult in complex scenarios. Power allocation also lacks a unified error measurement mechanism and a quantitative adjustment path for deviations from target values, resulting in resource waste, over-supply and under-supply of some loads during the power allocation phase, which can lead to system tripping and power module damage in severe cases. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the existing technology and to propose a heavy load query control method and system for a ship power management system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a heavy-load query control method for a ship power management system, comprising the following steps:

[0007] S1: Based on the remaining capacity of the ship's power module, the operating signals of critical load units and non-critical load units, the capacity range is divided and the boundaries are marked. The signal state combinations are compared to generate a power operation status table.

[0008] S2: Based on the power operation status table, call the access sequence controller to execute access action delay action reduction action, retrieve capacity range and load signal and sample multiple state transition sequences, use Monte Carlo algorithm to accumulate and average the reward of each action in the corresponding capacity range, obtain the benefit value of each action and write it into the table to generate a load action value matrix.

[0009] S3: Based on the load action numerical matrix, the corresponding unit is updated after accumulating and comparing the benefit value in each state. Threshold difference judgment is performed and the convergence state is marked. A greedy algorithm is used to select actions according to the benefit value in each convergence state and write them into the sequence in sequence. The adapted actions in all states are output to obtain the adapted action scheduling sequence.

[0010] S4: Based on the adaptive action scheduling sequence, call the access sequence controller to arrange the access sequence and calculate the power ratio. If the capacity limit is exceeded, correct and delete the corresponding item to form a power execution sequence table.

[0011] S5: Based on the power execution sequence list, establish the row and column mapping of the power allocation matrix, calculate the total error by summing the squared difference between the demand power and the allocated power, perform threshold judgment, stop if it is less than the threshold, and correct and update if it is greater than the threshold, and obtain the converged power allocation matrix.

[0012] As a further embodiment of the present invention, the power operation status table includes capacity range records, critical load status records, and non-critical load status records; the load action value matrix includes a capacity range index, an action type set, and a benefit value set; the adaptive action scheduling sequence includes a status number, an action instruction, and a benefit value; the power execution sequence table includes a load access order, a power allocation ratio, and a delay time parameter; and the convergent power allocation matrix includes a power supply row index, a load column index, and an allocated power value.

[0013] As a further aspect of the present invention, the specific steps for generating the power operation status table are as follows:

[0014] Based on the remaining capacity value of the ship's power module, the operating signals of critical load units and non-critical load units, the remaining capacity value is segmented and written into a table, the boundary values ​​of each interval are marked, the switching signals of critical load units and non-critical load units are recorded, and a capacity signal segmentation table is generated.

[0015] Based on the capacity signal segmentation table, the capacity range is compared with the load signal, the critical load unit signal and the non-critical load unit signal are matched, the capacity range identifier corresponding to the combination result is recorded, and the power operation status table is generated.

[0016] As a further aspect of the present invention, the specific steps for generating the load action numerical matrix are as follows:

[0017] Based on the power operation status table, capacity range data is retrieved and arranged in range order. The upper and lower limits of each range are marked. Combined with the signals of critical load units and non-critical load units, each item is written into the corresponding cell to generate a capacity signal matching table.

[0018] Based on the capacity signal matching table, the access sequence controller is called to sequentially allocate access action delay action reduction action, write the three types of actions into the capacity interval unit one by one, and mark the correspondence between the action number and the signal combination to generate an action status record table.

[0019] Based on the action status record table, the Monte Carlo algorithm is used to read the action number of each interval and assign a benefit value. The benefit value is written into the action cell, and a capacity interval index and value correspondence relationship is established to generate a load action value matrix.

[0020] As a further aspect of the present invention, the Monte Carlo algorithm first reads the action number corresponding to each capacity interval from the action state record table, performs multiple random samplings within the capacity interval for each action number, generates multiple state transition sequences, calculates the benefit results of the action under different sampling conditions during the execution of the sequence, accumulates all benefit results item by item and calculates the average value to obtain the benefit value of the action within the corresponding capacity interval, writes the benefit value into the action cell, and establishes a correspondence with the capacity interval index to form a load action value matrix.

[0021] As a further aspect of the present invention, the specific steps for generating the adapted action scheduling sequence are as follows:

[0022] Based on the load action value matrix, the benefit value is extracted for each state and added to the value of the previous round. The updated cumulative value is written to the corresponding state unit and archived according to the state number to generate a state benefit cumulative table.

[0023] Based on the state benefit cumulative table, the cumulative value of each state is extracted and compared with the difference of the previous round of records. States with a difference less than the threshold are marked as convergent states and written into the convergence identifier table to generate a convergence state identifier table.

[0024] Based on the convergence state label table, a greedy algorithm is used to read all converged states and extract the corresponding actions, assemble a complete sequence according to the state number, and write it into the action-state mapping list to obtain the adapted action scheduling sequence.

[0025] As a further aspect of the present invention, the greedy algorithm first reads all states marked as converged in the convergence state label table, extracts the corresponding benefit values ​​and candidate actions one by one, and then traverses them in order of state number. In each state, actions are selected according to the magnitude of the benefit value, always selecting the benefit action corresponding to the current state as the suitable action, and forming a mapping relationship with the corresponding state. This mapping relationship is written into the action-state mapping list, and no further backtracking or adjustment is performed. By processing all converged states in sequence, a complete action sequence is gradually constructed, forming a suitable action scheduling sequence, thereby realizing the determination and combination of suitable actions in all states.

[0026] As a further aspect of the present invention, the specific steps for generating the power execution sequence table are as follows:

[0027] Based on the adaptive action scheduling sequence, the access order is arranged, key load units are placed at the front and written into the table in sequence, the power ratio of each load is allocated, the delay time value is marked, and an access allocation sequence table is generated.

[0028] Based on the access allocation sequence table, the capacity limit is determined, the total power is calculated and compared with the power supply capacity limit, the excess value is corrected and the excess sequence item is deleted, the correction result is written into the table, and a power execution sequence table is generated.

[0029] As a further aspect of the present invention, the specific steps for generating the convergent power allocation matrix are as follows:

[0030] Based on the power execution sequence table, map the corresponding rows of the power module and the corresponding columns of the load unit, fill in the power allocation value at the intersection, calculate the difference between the power demand of each load and the allocated power and write it into the matrix unit to generate a power difference matrix.

[0031] Based on the power difference matrix, the squared differences are calculated and the total error value is obtained by summing them one by one. The error threshold is determined, and the matrix is ​​updated for the allocation values ​​that exceed the threshold. The operation is repeated until the condition is met, and the converged power allocation matrix is ​​obtained.

[0032] A heavy-load query control system for a ship power management system, the system being used to execute the aforementioned heavy-load query control method for the ship power management system, the system comprising:

[0033] Capacity segmentation module: Based on the remaining capacity value of the ship's power module, the capacity range boundary is set, and the operation signals of critical load units and non-critical load units are combined to form a combination to generate an operation capacity status mapping table.

[0034] State evaluation module: Based on the operating capacity state mapping table, it calls the access sequence controller's access, delay, and reduction action commands, samples the state transition path, uses the Monte Carlo algorithm to accumulate the response values ​​of each action in different capacity ranges, and generates an action response evaluation matrix.

[0035] Action generation module: Based on the action response evaluation matrix, calculate the difference in response values ​​between each state, filter convergent states, call a greedy algorithm to extract the action corresponding to the maximum response value and sort them, and generate an instruction adaptation execution sequence;

[0036] Power regulation module: Based on the instruction adaptation execution sequence, calculate the power of all connected loads, determine whether they exceed the limit, delete the exceeding items and renumber them, and generate a power allocation execution sequence list;

[0037] Energy convergence module: Based on the power allocation execution sequence list, establish a power mapping matrix, calculate the sum of squares of the difference between demand and allocated power, determine whether the threshold is met, iteratively correct, and obtain the dynamic energy allocation matrix.

[0038] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0039] In this invention, the Monte Carlo algorithm is introduced to construct a state transition structure, state behavior samples are set within the capacity range, transition paths under the combination of actions and capacity are sampled, capacity consumption response under each action is accumulated, and an action benefit matrix with quantifiable weights is generated by averaging the samples, forming a data-driven load response model and improving the clarity of scheduling action evaluation.

[0040] In this invention, after traversing the entire state space, the magnitude of the change in action reward is identified, and the structure is judged to converge by the difference threshold. Based on the convergence of the state, a greedy algorithm is called to read the action with the greatest benefit, and the action scheduling sequence is constructed in order according to the state number to avoid repeated selection of inefficient behavior.

[0041] In this invention, a hierarchical allocation mechanism is established by prioritizing access to critical loads and delaying and eliminating non-critical loads through access sorting and power ratio configuration logic. After establishing the power matrix structure, an error loop iteration mechanism is introduced. When the total error exceeds the upper limit, the allocation value is adjusted item by item. Through continuous correction until the error converges, the symmetry and fit of power configuration are improved, the action mismatch rate and allocation deviation rate are reduced, and the power scheduling convergence efficiency is improved. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0043] Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Example 1

[0045] Please see Figure 1 This invention provides a technical solution: a heavy-load query control method for a ship power management system, comprising the following steps:

[0046] S1: Based on the remaining capacity of the ship's power module, the operating signals of critical load units and non-critical load units, the capacity range is divided and the boundaries are marked. The signal state combinations are compared to generate a power operation status table.

[0047] S2: Based on the power operation status table, the access sequence controller is called to execute access action delay action reduction action. By retrieving capacity range and load signal and sampling multiple state transition sequences, the Monte Carlo algorithm is used to accumulate and average the reward of each action in the corresponding capacity range, obtain the benefit value of each action and write it into the table to generate the load action value matrix.

[0048] S3: Based on the load action numerical matrix, the corresponding unit is updated after accumulating and comparing the benefit value in each state. The threshold difference is judged and the convergence state is marked. The greedy algorithm is used to select the action according to the benefit value in each convergence state and write it into the sequence in sequence. The adapted actions in all states are output to obtain the adapted action scheduling sequence.

[0049] S4: Based on the adaptive action scheduling sequence, the access sequence controller is called to arrange the access order and calculate the power ratio. If the capacity limit is exceeded, the corresponding item is corrected and deleted to form a power execution sequence table.

[0050] S5: Based on the power execution sequence list, establish the row and column mapping of the power allocation matrix, calculate the total error by summing the squared difference between the demand power and the allocated power, perform threshold judgment, stop if it is less than the threshold, and correct and update if it is greater than the threshold, and obtain the converged power allocation matrix.

[0051] The power operation status table includes capacity range records, critical load status records, and non-critical load status records. The load action value matrix includes capacity range index, action type set, and benefit value set. The adaptive action scheduling sequence includes status number, action command, and benefit value. The power execution sequence table includes load access order, power allocation ratio, and delay time parameter. The convergent power allocation matrix includes power source row index, load column index, and allocated power value.

[0052] The specific steps for generating the power operation status table are as follows:

[0053] Based on the remaining capacity value of the ship's power module, the operating signals of critical load units and non-critical load units, the remaining capacity value is segmented and written into a table, the boundary values ​​of each interval are marked, the switching signals of critical load units and non-critical load units are recorded, and a capacity signal segmentation table is generated.

[0054] Based on the capacity signal segmentation table, the capacity range is compared with the load signal, the critical load unit signal and the non-critical load unit signal are matched, the capacity range identifier corresponding to the combination result is recorded, and the power operation status table is generated.

[0055] Based on the remaining capacity values ​​of the ship's power modules, the operating signals of key load units, and the operating signals of non-key load units, the remaining capacity values ​​are divided into intervals using an equidistant interval division method. The starting capacity is set to 0, the ending capacity to 100, and the step size to 10. A for loop is used to sequentially compare each capacity value with the upper and lower boundaries of each interval, and the capacity values ​​that meet the conditions are written to the field "CapacityIndex". At the same time, the lower and upper limits of each interval boundary are recorded and written to the fields "LowerBound" and "UpperBound". The switching signals of key load units and non-key load units are extracted and written to the status fields "KeySignalStatus" and "NonKeySignalStatus" in order of position number, generating a capacity signal segmentation table.

[0056] Based on the capacity signal segmentation table, a combined state mapping method is used to compare the capacity range with the load signal. A two-level nested loop structure is constructed. The outer layer traverses all capacity range identifiers, and the inner layer extracts the switch Boolean states from the signals of key load units and non-key load units. The two types of signals are concatenated using the position index to form a state combination code, which is written into the field "StateGroupCode". Then, each state combination code is associated with the current capacity range identifier in two fields and written into the structure unit. The data is summarized row by row to form a complete state set and generate a power operation status table.

[0057] The specific steps for generating the load action numerical matrix are as follows:

[0058] Based on the power operation status table, capacity range data is retrieved and arranged in range order. The upper and lower limits of each range are marked. Combined with the signals of critical load units and non-critical load units, the data is written into the corresponding cells one by one to generate a capacity signal matching table.

[0059] Based on the capacity signal matching table, the access sequence controller is called to allocate access action delay action reduction action in sequence. The three types of actions are written into the capacity interval unit one by one, and the action number and signal combination correspondence are marked to generate an action status record table.

[0060] Based on the action status record table, the Monte Carlo algorithm is used to read the action number of each interval and assign a benefit value. The benefit value is written into the action cell, and a capacity interval index and value correspondence relationship is established to generate a load action value matrix.

[0061] Based on the power operation status table, the capacity range field in the table is rearranged. The bubble sort method is used to use the start and end boundaries of each capacity range data as sorting keys. All ranges are compared and swapped one by one according to the start boundary value from smallest to largest. After sorting, the field identifier is called to renumber each record. The sorted upper and lower limits are written into the fields "LowerBound" and "UpperBound" respectively. The key load unit signals and non-key load unit signals in each range are extracted and written into the corresponding cells according to the Boolean expression of the status value to generate a capacity signal matching table.

[0062] Based on the capacity signal matching table, the access sequence controller is called to execute the access action, delay action, and reduction action numbering in sequence. A fixed order injection strategy is adopted, and the three types of actions are assigned action identifiers "ACT1", "ACT2", and "ACT3" respectively. The action identifiers are written into the cell where the capacity range is located one by one through the loop control structure, and the mapping relationship with the signal combination is marked in each row. The action number is recorded in the field "ActionCode" and the signal group code is recorded in the field "SignalGroup", generating an action status record table.

[0063] Based on the action status record table, the Monte Carlo algorithm is used to perform a reward assignment operation on the action number recorded in each capacity interval. The initial sample size is set to 500. An independent repeating sample sequence is generated for each action number. The action number and the corresponding capacity value are recorded in each sample sequence. The expected value is obtained by accumulating the difference between the corresponding capacity value of the action number and dividing it by the total number of samples. The expected value is written as the benefit value into the field "RewardScore". Then, a key-value structure is constructed according to the action number and the capacity interval number and written into the two-dimensional table cell to generate the load action value matrix.

[0064] The Monte Carlo algorithm first reads the action number corresponding to each capacity interval from the action state record table. For each action number, it performs multiple random samplings within the capacity interval to generate multiple state transition sequences. During the sequence execution, it calculates the benefit results of the action under different sampling conditions, accumulates all benefit results item by item and calculates the average value to obtain the benefit value of the action within the corresponding capacity interval. Then, it writes the benefit value into the action cell and establishes a correspondence with the capacity interval index to form a load action value matrix.

[0065] The Monte Carlo algorithm, according to the formula:

[0066] ;

[0067] in: Indicates the action number The expected overall benefit under all capacity state simulations. This indicates the first entry in the action status record table of the ship's power management system. The query action number, The first power distribution capacity represents the A discrete state interval, This represents the total number of samples taken for each action number and capacity interval combination in the Monte Carlo simulation. Indicates capacity status Next action The simulated original unit benefit value, Indicates the action number Historical performance stability indicators Indicates capacity state range The tension index Indicates action With capacity status The adaptation deviation value between them Represents the original benefit value item The weighting coefficients, Indicates historical stability The weighting coefficients, Indicating the quantity stress item The weighting coefficients, Indicates motion adaptation deviation. Weighting coefficients;

[0068] Execution process: First, based on the action status record table built into the ship's power management system, extract all heavy load query action numbers. And divide the capacity range according to the current system load status. This forms a capacity state set, representing the resource availability of each power subsystem within a specific time window. Then, the total number of rounds of the Monte Carlo simulation is set. For each group in turn The simulation is performed using a combination of random sampling methods. In each round, the original benefit function is first calculated. This is used to represent the unit power benefit that an action can achieve under the current capacity state. Then, historical task scheduling logs are extracted, and the stability function of the action number is calculated. The reliability index is obtained by back-calculating the output fluctuation variance of the action in multiple past power control responses, and then the stress level of the capacity range is assessed. Based on the statistical results of system resource scheduling, the scheduling frequency and resource occupancy ratio of capacity intervals are quantified. A two-vector model is constructed according to the action requirement configuration and capacity interval characteristics. The Euclidean distance is calculated and normalized to obtain the adaptation deviation function. Finally, the four values ​​are multiplied by the weighting coefficients. The weighting coefficients are obtained by performing multivariate regression modeling on the historical scheduling sample set and normalizing the sensitivity of each variable, thus completing the combined benefit evaluation of each round of sampling and accumulating the results. The results are then averaged to obtain the expected comprehensive benefit value of each action number across all capacity state intervals. It generates a load action numerical matrix to support the optimal power response strategy of the ship during heavy load control.

[0069] The specific steps for generating the adaptive action scheduling sequence are as follows:

[0070] Based on the load action numerical matrix, the benefit value is extracted for each state and added to the value of the previous round. The updated cumulative value is written to the corresponding state unit and archived according to the state number to generate a state benefit cumulative table.

[0071] Based on the state benefit cumulative table, the cumulative value of each state is extracted and compared with the difference of the previous round of records. States with a difference less than the threshold are marked as convergent states and written into the convergence label table to generate a convergence state label table.

[0072] Based on the convergence state label table, a greedy algorithm is used to read all converged states and extract the corresponding actions, assemble a complete sequence according to the state number, and write it into the action-state mapping list to obtain the adapted action scheduling sequence.

[0073] Based on the load action value matrix, the cell corresponding to the state number in each row of the matrix is ​​read to extract the benefit value of the current round. The benefit is superimposed using the state accumulation iteration method. The current value is added to the stored value of the previous round item by item. A loop structure is used to control the alignment of the state numbers and locate the corresponding cell positions. After accumulation, the new value is written to the cell corresponding to the field "AccValue". Then, the current state number and the corresponding updated value are written to the structure array in index order and output as a two-dimensional table to generate a state benefit accumulation table.

[0074] Based on the cumulative state benefit table, each value in the "AccValue" field is extracted and compared with the state value of the same number in the previous round. An error convergence identification method is used to calculate the absolute value of the difference in the order of state number. The difference result is compared with the set threshold of 5.0 item by item. If the current value is less than or equal to 5.0, the state is marked as "1" and if it is greater than 5.0, it is marked as "0". The results are written into the "ConvFlag" field and sorted by number to be output into a table structure to generate a convergence state mark table.

[0075] Based on the convergence state flag table, a greedy algorithm is used to perform action extraction operations on convergence states marked as "1". An index list is built for state numbers with "ConvFlag" as 1. The action values ​​corresponding to the numbers in the preceding structure are read sequentially. The maximum value in the "ActionCode" field is selected as the action for the state by using a conditional statement. The action number and the state number are combined into a key-value pair and written into a dictionary structure. All key-value pairs are then output as a list and written into the table field "MapActionToState" to obtain the adapted action scheduling sequence.

[0076] The greedy algorithm first reads all the states marked as convergent in the convergence state label table, extracts the corresponding benefit values ​​and candidate actions one by one, and then traverses them in order of state number. In each state, the action is selected according to the benefit value. The benefit action corresponding to the current state is always selected as the appropriate action, and a mapping relationship is formed with the corresponding state. This is written into the action-state mapping list, and no further backtracking or adjustment is performed. By processing all convergent states in sequence, a complete action sequence is gradually constructed, forming an appropriate action scheduling sequence, which realizes the determination and combination of appropriate actions in all states.

[0077] Greedy algorithm, according to the formula:

[0078] ;

[0079] in: Indicates scheduling to the 1st The system accumulates scheduling costs for each action. Indicates the first overload query The converged state number. This indicates the convergence state number of the previous executed step in the scheduling sequence. This indicates the jump distance between the current state and the previous state in terms of state number. Representation and State The associated action number, Indicates the execution of an action Required control response time Indicates action resource dependence Representing state The degree of emergency dispatch, This represents the weighting coefficient of the state jump distance in the cost function. This represents the weighting coefficient of the action response time in the cost function. This represents the weighting coefficient of action resource dependency in the cost function. This represents the weighting coefficient of the urgency of the state in the cost function;

[0080] Execution process: First, read the built-in convergence state flag table and extract all state numbers marked as converged. This indicates that the overload query action corresponding to the state has completed power characteristic modeling and can be used for scheduling control. Then, for each state... Bind to the corresponding action number This forms a set of actions to be scheduled, followed by initializing the action scheduling sequence and setting the target scheduling length. In each scheduling step, based on the current state number With candidate state jump distance Calculate the cost of state transition and simultaneously look up the action in the table. Average response time This is used to reflect the execution time consumption of an action, and then the resource coupling strength of the action in the current heavy-load control task is extracted through the system resource dependency graph. At the same time, assess the scheduling urgency of the current situation within the ship load window. If the state is in the critical load zone and there is a risk of load imbalance, a higher urgency weight is assigned, and then the four indicators are multiplied by their corresponding weight coefficients. The results are then summed to obtain the total scheduling cost of the current action in the given state. In each round, the state-action combination with the minimum scheduling cost is selected and written into the scheduling sequence. This process is repeated until the next round. Until all actions are sorted, an optimized action scheduling sequence is output, which is optimized in multiple dimensions such as state continuity, time consumption, resource dependence and response urgency, to support the intelligent response and precise control of the ship's power system in heavy load scenarios.

[0081] The specific steps for generating the power execution sequence table are as follows:

[0082] Based on the adaptive action scheduling sequence, the access order is arranged, key load units are placed at the front and written into the table in sequence, the power ratio of each load is allocated, the delay time value is marked, and an access allocation sequence table is generated.

[0083] Based on the access allocation sequence table, determine the capacity limit, calculate the total power and compare it with the power supply capacity limit, correct the excess value and delete the excess sequence item, write the correction result into the table, and generate the power execution sequence table.

[0084] Based on the adaptive action scheduling sequence, a priority sorting algorithm is used to perform a reordering operation on all load numbers in the scheduling sequence. The priority value of critical load units is set to 1, and the priority value of non-critical load units is set to 2. Nested conditional statements are used to identify and assign values ​​to the corresponding types of each number. Then, all numbers are sorted in ascending order of priority value, and the sorting results are written to the field "SortedIndex" item by item. The power ratio value is extracted from the load type field of each number and written to the field "PowerRatio". The default delay time is set to 5 seconds, and the delay time value of each item is written to the field "DelayTime" to generate the access allocation sequence table.

[0085] Based on the access allocation sequence table, the capacity is determined using the total power limit judgment method. The capacity limit is set to 100. All power ratio values ​​in the "PowerRatio" field are added sequentially and written to the "TotalPower" field. A single conditional statement is used to compare "TotalPower" with the limit value. If the current value is greater than the capacity limit, the sequence is rearranged from high to low according to the power ratio, and the accumulation operation is performed from the maximum value downward. If the accumulated sum is found to exceed the limit during the accumulation process, the corresponding load number is marked as "Exceed". All marked numbers are removed from the sequence, and the remaining numbers are written to the "ValidExecutionOrder" field in the original order to generate the power execution sequence table.

[0086] The specific steps for generating the convergent power allocation matrix are as follows:

[0087] Based on the power execution sequence table, the corresponding rows of the power modules and the corresponding columns of the load units are mapped, the power allocation values ​​at the intersections are filled in, the difference between the power demand of each load and the allocated power is calculated and written into the matrix unit to generate a power difference matrix.

[0088] Based on the power difference matrix, the square of the difference is calculated and the total error value is obtained by summing the values ​​one by one. The error threshold is determined, and the matrix is ​​updated for the allocation values ​​that exceed the threshold. The operation is repeated until the condition is met, and the converged power allocation matrix is ​​obtained.

[0089] Based on the power execution sequence list, a matrix mapping algorithm is used to establish the correspondence between power modules and load units. The matrix row labels are set as power module numbers and the column labels are as load unit numbers. In a double nested loop, a one-to-one mapping operation is performed on each module and load combination. The number in the field "ValidExecutionOrder" is read in the mapping order, and the corresponding allocated power value is extracted from the field "PowerRatio" and written into the cross cell. At the same time, the target power value in the field "RequiredPower" is read from the independent power demand dataset, and the difference operation is performed with the allocated power value. The result of the operation is written into the cell where the field "PowerDiff" is located, and the complete structure is output to form a power difference matrix.

[0090] Based on the power difference matrix, an error convergence iterative algorithm is used to square all differences and perform total error statistics. The error threshold is set to 20. A single-level traversal structure is used to read all values ​​in the field "PowerDiff" sequentially, perform the square operation, and write them to the field "SquareError". All squared values ​​are accumulated and written to the variable "TotalError". Then, a conditional statement is used to compare the value of "TotalError" with the threshold value. If it is greater, the cell with the largest value in the field "SquareError" is located, the cell corresponding to the allocated power value is traced back, and the downward adjustment operation is performed. By default, the downward adjustment ratio in each round is 10% of the current value. The updated and modified value is written back to the original position, the difference is recalculated, and "PowerDiff" is updated. The above operation is repeated until "TotalError" is less than or equal to the threshold, and the converged power allocation matrix is ​​obtained.

[0091] Please see Figure 2 A heavy-load query control system for a ship power management system, the system being used to execute the aforementioned heavy-load query control method for the ship power management system, the system comprising:

[0092] Capacity segmentation module: Based on the remaining capacity value of the ship's power module, the capacity range boundary is set, and the operation signals of critical load units and non-critical load units are combined to form a combination to generate an operation capacity status mapping table.

[0093] State evaluation module: Based on the running capacity state mapping table, it calls the access sequence controller's access, delay, and reduction action commands, samples the state transition path, uses the Monte Carlo algorithm to accumulate the response values ​​of each action in different capacity ranges, and generates an action response evaluation matrix.

[0094] Action generation module: Based on the action response evaluation matrix, calculate the difference in response values ​​between states, filter convergent states, call a greedy algorithm to extract the action corresponding to the maximum response value and sort it, and generate an instruction adaptation execution sequence;

[0095] Power regulation module: Based on the instruction adaptation execution sequence, calculate the power of all connected loads, determine whether they exceed the limit, delete the exceeding items and renumber them, and generate a list of power allocation execution sequences;

[0096] Energy convergence module: Based on the power allocation execution sequence list, a power mapping matrix is ​​established, the sum of squares of the difference between demand and allocated power is calculated, the threshold is determined, iterative correction is performed, and the dynamic energy allocation matrix is ​​obtained.

[0097] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A heavy-load interrogation control method for a ship power management system, characterized in that, Includes the following steps: S1: Based on the remaining capacity of the ship's power module, the operating signals of critical load units and non-critical load units, the capacity range is divided and the boundaries are marked. The signal state combinations are compared to generate a power operation status table. S2: Based on the power operation status table, call the access sequence controller to execute access action delay action reduction action, retrieve capacity range and load signal and sample multiple state transition sequences, use Monte Carlo algorithm to accumulate and average the reward of each action in the corresponding capacity range, obtain the benefit value of each action and write it into the table to generate a load action value matrix. S3: Based on the load action numerical matrix, the corresponding unit is updated after accumulating and comparing the benefit value in each state. Threshold difference judgment is performed and the convergence state is marked. A greedy algorithm is used to select actions according to the benefit value in each convergence state and write them into the sequence in sequence. The adapted actions in all states are output to obtain the adapted action scheduling sequence. S4: Based on the adaptive action scheduling sequence, call the access sequence controller to arrange the access sequence and calculate the power ratio. If the capacity limit is exceeded, correct and delete the corresponding item to form a power execution sequence table. S5: Based on the power execution sequence list, establish the row and column mapping of the power allocation matrix, calculate the total error by summing the squared difference between the demand power and the allocated power, perform threshold judgment, stop if it is less than the threshold, and correct and update if it is greater than the threshold, and obtain the converged power allocation matrix.

2. The heavy-load query control method for a ship power management system according to claim 1, characterized in that, The power operation status table includes capacity range records, critical load status records, and non-critical load status records. The load action value matrix includes a capacity range index, an action type set, and a benefit value set. The adaptive action scheduling sequence includes a status number, action command, and benefit value. The power execution sequence table includes a load access order, power allocation ratio, and delay time parameter. The convergent power allocation matrix includes a power supply row index, a load column index, and allocated power value.

3. The heavy-load query control method for a ship power management system according to claim 1, characterized in that, The specific steps for generating the power operation status table are as follows: Based on the remaining capacity value of the ship's power module, the operating signals of critical load units and non-critical load units, the remaining capacity value is segmented and written into a table, the boundary values ​​of each interval are marked, the switching signals of critical load units and non-critical load units are recorded, and a capacity signal segmentation table is generated. Based on the capacity signal segmentation table, the capacity range is compared with the load signal, the critical load unit signal and the non-critical load unit signal are matched, the capacity range identifier corresponding to the combination result is recorded, and the power operation status table is generated.

4. The heavy-load query control method for a ship power management system according to claim 1, characterized in that, The specific steps for generating the load action numerical matrix are as follows: Based on the power operation status table, capacity range data is retrieved and arranged in range order. The upper and lower limits of each range are marked. Combined with the signals of critical load units and non-critical load units, each item is written into the corresponding cell to generate a capacity signal matching table. Based on the capacity signal matching table, the access sequence controller is called to sequentially allocate access action delay action reduction action, write the three types of actions into the capacity interval unit one by one, and mark the correspondence between the action number and the signal combination to generate an action status record table. Based on the action status record table, the Monte Carlo algorithm is used to read the action number of each interval and assign a benefit value. The benefit value is written into the action cell, and a capacity interval index and value correspondence relationship is established to generate a load action value matrix.

5. The heavy-load query control method for a ship power management system according to claim 4, characterized in that, The Monte Carlo algorithm first reads the action number corresponding to each capacity interval from the action state record table. For each action number, it performs multiple random samplings within the capacity interval to generate multiple state transition sequences. During the execution of the sequences, it calculates the benefit results of the action under different sampling conditions, accumulates all benefit results item by item and calculates the average value to obtain the benefit value of the action within the corresponding capacity interval. Then, it writes the benefit value into the action cell and establishes a correspondence with the capacity interval index to form a load action value matrix.

6. The heavy-load query control method for a ship power management system according to claim 1, characterized in that, The specific steps for generating the adapted action scheduling sequence are as follows: Based on the load action value matrix, the benefit value is extracted for each state and added to the value of the previous round. The updated cumulative value is written to the corresponding state unit and archived according to the state number to generate a state benefit cumulative table. Based on the state benefit cumulative table, the cumulative value of each state is extracted and compared with the difference of the previous round of records. States with a difference less than the threshold are marked as convergent states and written into the convergence identifier table to generate a convergence state identifier table. Based on the convergence state label table, a greedy algorithm is used to read all converged states and extract the corresponding actions, assemble a complete sequence according to the state number, and write it into the action-state mapping list to obtain the adapted action scheduling sequence.

7. The heavy-load query control method for a ship power management system according to claim 6, characterized in that, The greedy algorithm first reads all the converged states in the convergence state label table, extracts the corresponding benefit values ​​and candidate actions one by one, and then traverses them in order of state number. In each state, the action is selected according to the benefit value, always selecting the benefit action corresponding to the current state as the suitable action, and forming a mapping relationship with the corresponding state. This mapping relationship is written into the action-state mapping list, and no further backtracking or adjustment is performed. By processing all converged states in sequence, a complete action sequence is gradually constructed, forming a suitable action scheduling sequence, thereby realizing the determination and combination of suitable actions in all states.

8. The heavy-load query control method for a ship power management system according to claim 1, characterized in that, The specific steps for generating the power execution sequence table are as follows: Based on the adaptive action scheduling sequence, the access order is arranged, key load units are placed at the front and written into the table in sequence, the power ratio of each load is allocated, the delay time value is marked, and an access allocation sequence table is generated. Based on the access allocation sequence table, the capacity limit is determined, the total power is calculated and compared with the power supply capacity limit, the excess value is corrected and the excess sequence item is deleted, the correction result is written into the table, and a power execution sequence table is generated.

9. The heavy-load query control method for a ship power management system according to claim 1, characterized in that, The specific steps for generating the convergent power allocation matrix are as follows: Based on the power execution sequence table, map the corresponding rows of the power module and the corresponding columns of the load unit, fill in the power allocation value at the intersection, calculate the difference between the power demand of each load and the allocated power and write it into the matrix unit to generate a power difference matrix. Based on the power difference matrix, the squared differences are calculated and the total error value is obtained by summing them one by one. The error threshold is determined, and the matrix is ​​updated for the allocation values ​​that exceed the threshold. The operation is repeated until the condition is met, and the converged power allocation matrix is ​​obtained.

10. A heavy-load interrogation control system for a ship power management system, characterized in that, The heavy-load query control method for a ship power management system according to any one of claims 1-9, wherein the system comprises: Capacity segmentation module: Based on the remaining capacity value of the ship's power module, the capacity range boundary is set, and the operation signals of critical load units and non-critical load units are combined to form a combination to generate an operation capacity status mapping table. State evaluation module: Based on the operating capacity state mapping table, it calls the access sequence controller's access, delay, and reduction action commands, samples the state transition path, uses the Monte Carlo algorithm to accumulate the response values ​​of each action in different capacity ranges, and generates an action response evaluation matrix. Action generation module: Based on the action response evaluation matrix, calculate the difference in response values ​​between each state, filter convergent states, call a greedy algorithm to extract the action corresponding to the maximum response value and sort them, and generate an instruction adaptation execution sequence; Power regulation module: Based on the instruction adaptation execution sequence, calculate the power of all connected loads, determine whether they exceed the limit, delete the exceeding items and renumber them, and generate a power allocation execution sequence list; Energy convergence module: Based on the power allocation execution sequence list, establish a power mapping matrix, calculate the sum of squares of the difference between demand and allocated power, determine whether the threshold is met, iteratively correct, and obtain the dynamic energy allocation matrix.

Images