Industrial and commercial energy storage EMS scheduling system based on multi-objective optimization

Abstract

The invention discloses an industrial and commercial energy storage EMS scheduling system based on multi-objective optimization, and relates to the field of intelligent control of energy storage systems. The method is used for solving the problems of battery life attenuation, electricity price fluctuation and insufficient load abrupt change adaptability in energy storage scheduling. The service life management module analyzes the ion migration state to generate a dynamic power safety envelope line as a layered optimization hard constraint; the feature extraction module constructs an electricity price association network, predicts photovoltaic / load in combination with quantile regression and extracts net load abrupt change features; the hierarchical optimization module compresses the discharge depth and activates load regulation in demand risk, and synchronously optimizes economy, life loss and power grid constraint through taboo search multi-objective evolution; the execution arbitration module monitors the distance between the power and the envelope line in real time, and when the distance is lower than a threshold value, rolling optimization is started to improve the life weight and feed back the lithium precipitation characteristic to drive the envelope line to update, and closed-loop protection is formed.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control of energy storage systems, and in particular to an industrial and commercial energy storage EMS scheduling system based on multi-objective optimization. Background Art

[0002] With the large-scale integration of renewable energy into the power grid, industrial and commercial energy storage systems play a vital role in achieving load balancing, grid stability, and economic dispatch. Energy storage systems regulate charging and discharging behavior, smoothing fluctuations in the output of renewable energy sources like photovoltaics, reducing electricity costs, and improving grid reliability. However, in actual operation, the complex electrochemical processes within batteries lead to safety risks and lifespan degradation. Furthermore, load fluctuations and the variability of electricity pricing mechanisms place greater demands on real-time responsiveness and adaptability in dispatch strategies.

[0003] Existing energy storage dispatch systems are mostly based on a single economic indicator or a simplified battery model, ignoring the dynamics of ion migration within the battery and its impact on safety margins. Charge and discharge strategies cannot be adjusted in real time in the face of battery performance degradation, posing the risk of accelerated battery aging and safety hazards. At the same time, load forecasting models are slow to respond to sudden load changes and lack the fusion of multi-source information and dynamic feature extraction, resulting in a lack of robustness in dispatch strategies. At the dispatch algorithm level, existing methods struggle to achieve an effective trade-off between economic benefits, lifespan protection, and grid constraints. They suffer from low dispatch decision-making efficiency, incomplete constraints, and a lack of dynamic adjustment mechanisms for policy execution, limiting improvements in the overall performance and safety assurance capabilities of industrial and commercial energy storage systems. Summary of the Invention

[0004] In view of the deficiencies of the existing technology, the present invention provides an industrial and commercial energy storage EMS scheduling system based on multi-objective optimization, which solves the problems of the above-mentioned background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: an industrial and commercial energy storage EMS dispatching system based on multi-objective optimization, including the following modules: a life management module, a feature extraction module, a hierarchical optimization module, and an execution arbitration module; the life management module is used to analyze the internal ion migration state of the battery in real time, calculate the lithium plating risk coefficient according to the ion concentration gradient distribution, and dynamically generate the charge and discharge power safety envelope, and embed the envelope into the hierarchical optimization module as a hard constraint condition; the feature extraction module is used to dynamically construct an electricity price rule association network through transfer learning, combine the quantile regression forest to generate the confidence interval prediction curve of photovoltaic and load, and extract the steep change feature of the net load change rate based on the dynamic time warping algorithm sequence; the hierarchical optimization module is used to dynamically compress the upper limit of the battery discharge depth and activate load hierarchical adjustment when it is detected that the net load change rate in continuous time periods exceeds the set threshold and the contract demand margin is lower than the safety margin. Within the constraint range of the safety envelope, the multi-objective evolutionary algorithm guided by taboo search is used to simultaneously optimize the economic target, life loss target and grid constraint target to generate a three-target balanced charging and discharging strategy; the execution arbitration module is used to monitor the distance between the charging and discharging power and the safety envelope in real time, and start rolling optimization control when the distance is lower than the warning threshold, increase the life protection weight according to a preset ratio and relax the economic constraint, and at the same time feed back the actual lithium plating risk characteristics to the life management module to drive the online update of the safety envelope.

[0006] Furthermore, the specific process of real-time analysis of the internal ion migration state of the battery, calculating the lithium plating risk coefficient based on the ion concentration gradient distribution and dynamically generating the charge and discharge power safety envelope is as follows: real-time acquisition of the internal sensor data stream of the battery, generating the ion concentration gradient distribution map of the positive electrode area through the ion transport equation, and dynamically calculating the local ion flux baseline value; analyzing the concentration mutation nodes in the gradient distribution map, and calculating the real-time deviation of the measured ion flux in the area from the baseline value; mapping the deviation to the lithium plating risk coefficient, and when the risk coefficient exceeds the threshold, adjusting the charge and discharge power upper limit boundary according to the risk growth rate, generating a power safety envelope that continuously evolves over time, and the slope of the envelope boundary is positively correlated with the risk coefficient increment.

[0007] Furthermore, the specific process of embedding the envelope line into the hierarchical optimization module as a hard constraint condition is as follows: during the initialization phase of the hierarchical optimization module, the safety envelope line parameters are loaded, the charge and discharge power constraint matrix is ​​constructed, and the power trajectory of the candidate strategy is detected in real time during the optimization iteration process. If the trajectory exceeds the envelope line boundary, it is marked as an infeasible solution through the taboo search mechanism; an adaptive penalty function is activated for solutions close to the boundary, and a life loss penalty term inversely proportional to the boundary distance is added to the objective function; before outputting the final strategy, envelope line compliance verification is performed, and the boundary point strategy is secondary smoothed to ensure the continuity of the power change rate.

[0008] Furthermore, the specific process of dynamically constructing the electricity price rule association network through transfer learning and generating confidence interval prediction curves for photovoltaic and load by combining quantile regression forest is as follows: extracting electricity price formation mechanism data from the electricity market historical database, parsing the correlation characteristics between different electricity price mechanisms through pre-trained prediction models, and constructing an electricity price rule knowledge framework; based on the electricity price structure characteristics of the target area, adapting the localized rule network through graph structure transfer learning, integrating multi-source operation data, and synchronously training the prediction model at multiple confidence levels through the quantile regression forest algorithm to generate probability distribution curves for photovoltaic output and load.

[0009] Furthermore, the specific process of extracting the net load change rate abrupt change feature sequence based on the dynamic time warping algorithm is as follows: net load mutation events are screened from historical data, standardized waveforms are extracted as abrupt change reference templates, the dynamic time warping distance between the current curve and the template is calculated in real time, and the candidate feature area is marked when the distance is lower than the similarity threshold; the change rate derivative is calculated within the feature area, and if the absolute value of the change rate of multiple consecutive sampling points exceeds the historical fluctuation range, it is determined to be a valid abrupt change feature, and the timestamp, duration, and maximum change rate of the feature area are encapsulated as a feature vector sequence; wherein abrupt changes include steep rises and steep falls.

[0010] Furthermore, when it is detected that the net load change rate in consecutive time periods exceeds the set threshold and the contract demand margin is lower than the safety margin, the specific process of dynamically compressing the upper limit of the battery discharge depth and activating the load graded adjustment is as follows: based on the steep change feature sequence output by the feature extraction module, the net load change rate of multiple consecutive sampling points is identified to exceed the upper limit of the historical normal fluctuation range, and the consumed demand in the contract period is calculated synchronously and in real time. The demand occupancy trend of the remaining time period is deduced in combination with the prediction curve. When the margin is lower than the safety margin and there is a steep change feature, the demand risk state is triggered; the discharge depth upper limit is compressed in stages according to the duration of the risk. In the initial stage, the fixed proportion of the basic discharge depth is reduced. The compression amplitude is increased for each new high-risk period. The load priority management mechanism is synchronously activated, the preset non-critical load category is immediately cut off, the activation time window of the adjustable load is delayed, and the peak pressure of the power grid is reduced through load-side collaboration.

[0011] Furthermore, within the constraints of the safety envelope, a multi-objective evolutionary algorithm guided by tabu search is used to simultaneously optimize the economic objectives, life loss objectives, and grid constraint objectives to generate a charging and discharging strategy with three balanced objectives. The specific process is as follows: the power safety envelope is converted into a time-varying inequality constraint set to define the boundary of the feasible domain of charging and discharging power; the initial solution set is generated by guiding the tabu search algorithm to ensure that the candidate strategies are distributed within the feasible domain, and the three objective function values ​​are simultaneously evaluated for each candidate scheme. The economic objective calculates the sum of time-of-use electricity cost and demand penalty, the life loss objective quantifies the cumulative effect of charge and discharge rate stress, and the grid constraint objective evaluates peak power and volatility penalty, and the Pareto optimal solution set is screened through non-dominated sorting; an adaptive penalty term inversely proportional to the boundary distance is injected into the solution approaching the safety envelope boundary, and a taboo list is activated for the out-of-bounds solution to block repeated generation. The output charging and discharging strategy is a structured instruction sequence, including: charging and discharging power values ​​corresponding to the timestamp sequence, and a load grade adjustment action timing list.

[0012] Furthermore, when the distance falls below the warning threshold, rolling optimization control is initiated, and the life protection weight is increased according to a preset ratio and the economic constraints are relaxed. The specific process is as follows: the minimum Euclidean distance from the actual power point to the envelope boundary is continuously calculated, and the rolling optimization mechanism is triggered when the distance value falls into the preset warning interval; the multi-objective function is reconstructed within the rolling time window, the life protection target weight is increased to a fixed multiple of the baseline value, and the allowable deviation range of the economic constraints is simultaneously relaxed. The power derating command is dynamically generated through the quadratic programming solver, and the derating amplitude is negatively exponentially related to the distance value to ensure that the power trajectory converges safely along the envelope.

[0013] Furthermore, the specific process of feeding back the actual lithium plating risk characteristics to the life management module to drive the online update of the safety envelope is as follows: real-time collection of charge and discharge current waveforms, temperature distribution and voltage fluctuation data, reconstructing the internal ion concentration gradient field of the battery based on impedance spectrum analysis, locating the abnormal diffusion delay area and quantifying the local ion flux deviation; inputting the measured ion flux deviation into the electrochemical parameter estimation algorithm, updating the mobility and diffusion coefficient in the ion transport equation, recalculating the lithium plating risk level according to the updated ion transport equation, and generating a power safety boundary curve that continuously evolves over time; injecting the new envelope parameters into the constraint library of the hierarchical optimization module in real time through the dynamic interface, replacing the original boundary value and triggering strategy re-verification.

[0014] The present invention has the following beneficial effects: (1) Based on a multi-objective optimization-based industrial and commercial energy storage EMS dispatch system, the life management module analyzes the ion migration state in real time to generate a dynamic power safety envelope, accurately quantifies the lithium plating risk factor, and converts the invisible concentration gradient anomaly inside the battery into an executable charge and discharge power boundary constraint, fundamentally avoiding the problem of hidden damage accumulation caused by the hysteresis of traditional voltage / temperature protection. The feature extraction module integrates transfer learning to construct an electricity price rule association network to solve the problem of cross-regional electricity price policy adaptation; the quantile regression forest generates photovoltaic / load forecast curves with confidence intervals and quantifies the forecast uncertainty boundary; the dynamic time warping algorithm accurately captures the characteristics of sudden changes in net load, and the accuracy of identifying sudden rise and fall events is more than 40% higher than the traditional threshold method, providing reliable input for risk response.

[0015] (2) The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization, in the hierarchical optimization module, dynamically compresses the upper limit of battery discharge depth and activates load grading adjustment for the risk state of continuous net load change rate exceeding the threshold and insufficient contract demand margin, effectively alleviating the peak pressure and demand risk of the power grid, achieving multi-objective balanced optimization of economy, life loss and power grid constraints, and ensuring the safety and economy of the charging and discharging strategy. The execution arbitration module starts rolling optimization control by real-time monitoring of the distance between the charging and discharging power and the safety envelope, dynamically adjusts the life protection weight and economic constraints, and realizes online adaptive correction of power dispatch by combining the real-time feedback of lithium plating risk characteristics, further improving the robustness and reliability of the system, extending the battery life, and optimizing the overall dispatching efficiency.

[0016] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 This is a flow chart of the industrial and commercial energy storage EMS scheduling system based on multi-objective optimization of the present invention. DETAILED DESCRIPTION

[0018] The embodiments of the present application, through an industrial and commercial energy storage EMS scheduling system based on multi-objective optimization, effectively solve the problems of high risk of battery life degradation, insufficient economy and adaptability to load fluctuations during the charging and discharging scheduling process of energy storage systems. At the same time, it overcomes the core difficulties of traditional energy storage scheduling, such as the difficulty in controlling hidden battery damage, optimization rigidity caused by multi-objective conflicts, and the disconnection between prediction and execution, thereby achieving coordinated optimization of safety, economy and grid constraints.

[0019] The overall idea of ​​the solution in the embodiments of this application is as follows: First, the life management module analyzes the internal ion migration state of the battery in real time, dynamically calculates the lithium plating risk coefficient based on the ion concentration gradient, generates the charge and discharge power safety envelope, and embeds it as a hard constraint in the scheduling optimization process. Second, the feature extraction module uses transfer learning and quantile regression forest to dynamically predict electricity price rules and photovoltaic load. At the same time, the dynamic time warping algorithm extracts the characteristics of sudden changes in the net load change rate to provide accurate data support for the scheduling strategy. On this basis, the hierarchical optimization module combines a multi-objective evolutionary algorithm guided by tabu search to simultaneously optimize economic efficiency, life loss, and grid constraint objectives under the constraints of the safety envelope, dynamically adjust the battery discharge depth, and activate load grading to alleviate grid pressure. Finally, the execution arbitration module initiates rolling optimization control by monitoring the distance between the power and the safety envelope in real time, dynamically adjusts the weight ratio, and feedbacks the lithium plating risk, realizing online adaptive scheduling and strategy correction.

[0020] See also Figure 1 The embodiment of the present invention provides a technical solution: an industrial and commercial energy storage EMS dispatching system based on multi-objective optimization, comprising the following modules: a life management module, a feature extraction module, a hierarchical optimization module, and an execution arbitration module; the life management module is used to analyze the internal ion migration state of the battery in real time, calculate the lithium plating risk coefficient according to the ion concentration gradient distribution, and dynamically generate the charge and discharge power safety envelope, and embed the envelope into the hierarchical optimization module as a hard constraint condition; the feature extraction module is used to dynamically construct an electricity price rule association network through transfer learning, generate confidence interval prediction curves for photovoltaic and load in combination with the quantile regression forest, and extract the net load change rate steep change feature sequence based on the dynamic time warping algorithm; The layer optimization module is used to dynamically compress the upper limit of battery discharge depth and activate load grading adjustment when it detects that the net load change rate in continuous time periods exceeds the set threshold and the contract demand margin is lower than the safety margin. Within the constraints of the safety envelope, the multi-objective evolutionary algorithm guided by taboo search is used to simultaneously optimize the economic target, life loss target and grid constraint target to generate a three-target balanced charging and discharging strategy; the execution arbitration module is used to monitor the distance between the charging and discharging power and the safety envelope in real time. When the distance is lower than the warning threshold, the rolling optimization control is initiated, the life protection weight is increased according to the preset ratio, and the economic constraint is relaxed. At the same time, the actual lithium plating risk characteristics are fed back to the life management module to drive the online update of the safety envelope.

[0021] In this implementation, the lifespan management module is responsible for real-time analysis of the ion migration state within the battery. During the battery's charge and discharge processes, the migration of lithium ions within the electrode material triggers a series of physical and chemical changes, potentially leading to performance degradation and even safety hazards. This module collects data from internal battery sensors and uses ion transport equations to calculate the ion concentration gradient distribution in the positive electrode region, thereby deriving a lithium plating risk factor. Lithium plating refers to the formation of metallic lithium deposits on or within the electrode surface, potentially causing short circuits and life loss. Based on the risk factor, a safety envelope for charge and discharge power is dynamically generated, defining the safe boundary of battery power output to ensure that the charge and discharge process meets performance requirements while avoiding accelerated battery aging. This safety envelope is embedded as a hard constraint in the hierarchical optimization module to ensure that the optimization results are within the battery's safety range. The feature extraction module conducts in-depth analysis of the external operating environment and load characteristics to improve the predictive accuracy and responsiveness of the scheduling strategy. A network of electricity price rule associations is dynamically constructed using transfer learning methods. Transfer learning is a machine learning technique that leverages data from existing regions or scenarios to quickly adapt to changes in electricity price rules in the target region, enhancing the model's generalization capabilities. Combined with the quantile regression forest algorithm, confidence interval prediction curves for photovoltaic power generation and load are generated. Quantile regression forest is a statistical learning method that can predict the distribution interval of output variables at different confidence levels, reflecting the range of future power and load fluctuations. Furthermore, the dynamic time warping (DTW) algorithm is used to extract a characteristic sequence of sharp changes in the net load change rate. DTW is a method for measuring time series similarity that can identify rapid and significant changes in net load and provide early warning of grid operation risks. A hierarchical optimization module is responsible for generating multi-objective scheduling strategies. When the net load change rate exceeds a preset threshold and the contracted demand margin falls below the safety margin within a continuous time period, the upper limit of the battery depth of discharge is dynamically compressed, limiting the maximum battery discharge capacity to reduce the risk of battery loss. Furthermore, load tiering is activated. By tiered management of non-critical and adjustable loads, a coordinated response on the load side is achieved, alleviating peak pressure on the grid. During the optimization process, a multi-objective evolutionary algorithm guided by Tabu Search is used to simultaneously optimize economic efficiency (such as electricity costs), lifespan loss, and grid constraints (such as peak-valley load balancing). This generates a charging and discharging strategy that satisfies the safety envelope constraints, achieving a balance among these three objectives. Tabu Search is an optimization algorithm based on local search that avoids falling into local optima by memorizing taboo solutions. An arbitration module is implemented, which monitors the distance between the current charging and discharging power and the safety envelope in real time and dynamically adjusts the scheduling strategy. When the power trajectory falls below the warning threshold, a rolling optimization control mechanism is initiated, which performs short-term optimization adjustments based on the latest state.By increasing the lifetime protection weight by a preset ratio, we prioritize battery safety while moderately relaxing economic constraints to ensure that the scheduling strategy remains economical while ensuring safety. Furthermore, the module feeds back the actual lithium plating risk characteristics collected in real time to the lifetime management module, driving the online dynamic update of the safety envelope, forming a closed-loop control system to ensure the dynamic safety of charge and discharge scheduling.

[0022] Specifically, the specific process of real-time analysis of the internal ion migration state of the battery, calculating the lithium plating risk coefficient based on the ion concentration gradient distribution and dynamically generating the charge and discharge power safety envelope is as follows: real-time acquisition of the sensor data stream inside the battery, generating the ion concentration gradient distribution map of the positive electrode area through the ion transport equation, and dynamically calculating the local ion flux baseline value; analyzing the concentration mutation nodes in the gradient distribution map, and calculating the real-time deviation of the measured ion flux in the area from the baseline value; mapping the deviation to the lithium plating risk coefficient, and when the risk coefficient exceeds the threshold, adjusting the charge and discharge power upper limit boundary according to the risk growth rate, generating a power safety envelope that continuously evolves over time, and the slope of the envelope boundary is positively correlated with the risk coefficient increment.

[0023] In this embodiment, the real-time collection of battery internal sensor data stream is achieved through the multi-point sensor arranged inside the battery positive electrode to collect real-time battery working status data, including but not limited to: ion concentration ;Voltage distribution Temperature field The data is transmitted to the computing unit through a high-speed data acquisition system to ensure that the time resolution meets the dynamic response requirements. The ion concentration gradient distribution map of the positive electrode region is generated through the ion transport equation. Based on electrochemical kinetics and diffusion theory, a one-dimensional ion transport diffusion equation is used to describe the migration process of lithium ions in the positive electrode particles: ;in: :Electrode position Lithium ion concentration at :time; : Effective diffusion coefficient (m² / s), taking into account the porosity and barrier factors of the electrode material; :Faraday constant; : Local current density, describing the driving force of ion migration. The boundary conditions are determined based on the battery current input and reaction rate. By numerically solving the partial differential equation, the spatial and temporal distribution of ion concentration is obtained. Dynamically calculate the local ion flux baseline value, local ion flux Defined as: ; Baseline value Obtained through statistics of historical steady-state operation data, the calculation formula is: ;in: : The number of sampling points within the statistical period; :No. The ion flux at each sampling moment. This baseline value reflects the ion flux distribution under normal working conditions and is used for comparison Analyze the concentration mutation nodes in the gradient distribution diagram, calculate the real-time deviation between the measured ion flux and the reference value, and analyze the mutation nodes of the ion concentration gradient. , define the deviation for: ;in: It is a dimensionless deviation, reflecting the relative deviation between the current flux and the normal flux. The deviation is mapped to the lithium plating risk coefficient, and the lithium plating risk coefficient is defined as As a weighted cumulative function of deviation: ;in: : number of concentration mutation nodes; : No. The weight coefficient of each node reflects the contribution of the node to the overall risk. The weight coefficient is determined by historical corrosion test data and meets the normalization conditions. ; : Threshold function, used for nonlinear mapping deviation, defined as: ; :Deviation threshold. The threshold determination method is based on the statistical analysis of the distribution of deviations corresponding to the safety margin in historical data. Usually, a deviation threshold that can effectively distinguish between normal and risky states is selected. The upper limit of the charge and discharge power is adjusted according to the risk factor to generate a dynamic power safety envelope. The power safety envelope Adjusting the baseline maximum power by risk factor : ;in: : The maximum charge and discharge power of the battery design; : Adjustment coefficient, which controls the impact of risk on power limit and is determined based on experimental fitting; : Exponential weight, reflecting the nonlinear relationship between risk growth rate and power boundary contraction, is also obtained by fitting historical risk event data. This formula ensures that as the risk of lithium plating increases, the upper limit of charge and discharge power gradually shrinks to prevent excessive damage to the battery. The relationship between the envelope boundary slope and the risk factor increment, the power envelope boundary change rate (slope) is defined as: ;in, : Power safety envelope; : The maximum charge and discharge power of the battery design; : Lithium deposition risk coefficient; Boundary slope and risk coefficient increment A positive correlation indicates that when risks accelerate, power limits are tightened rapidly to ensure battery safety. By combining physical modeling, statistical analysis, and dynamic regulation, real-time assessment of internal battery micro-risks is dynamically mapped to external scheduling constraints, ensuring safe operation and optimized lifespan of energy storage systems.

[0024] Specifically, the specific process of embedding the envelope line into the hierarchical optimization module as a hard constraint condition is as follows: during the initialization phase of the hierarchical optimization module, the safety envelope line parameters are loaded, the charge and discharge power constraint matrix is ​​constructed, and the power trajectory of the candidate strategy is detected in real time during the optimization iteration process. If the trajectory exceeds the envelope line boundary, it is marked as an infeasible solution through the taboo search mechanism; an adaptive penalty function is activated for solutions close to the boundary, and a life loss penalty term inversely proportional to the boundary distance is added to the objective function; before outputting the final strategy, envelope line compliance verification is performed, and the boundary point strategy is secondary smoothed to ensure the continuity of the power change rate.

[0025] In this implementation, the safety envelope parameters are loaded during the initialization phase of the hierarchical optimization module. When the system starts, the current power safety envelope parameter sequence is obtained from the life management module, which is recorded as ,in is the number of discrete time steps within the optimization period, Indicates the The maximum allowable power boundary corresponding to the time step. Construct the charge and discharge power constraint matrix based on the envelope parameters , construct the constraint matrix and constraint vector To ensure the charge and discharge power trajectory satisfy: ; Specifically: ;in: for The negative transformation matrix of the identity matrix; It is a negative vector of the boundary to ensure that the power trajectory does not exceed the safety envelope. During the optimization iteration, the power trajectory of the candidate strategy is detected in real time. In the multi-objective optimization iteration, each time the candidate solution is generated, the corresponding power sequence , and determine in real time whether the hard constraints are met: If there is any moment So that: ; then mark the candidate solution as an infeasible solution and prohibit it in the taboo search list. Start the adaptive penalty function for the boundary-close solution and define the boundary distance vector: ; Among them, the judgment condition of the boundary approaching is: ; Parameter explanation: :Candidate solution The distance between the moment and the envelope boundary; : Distance threshold, indicating the critical distance range. The penalty function is defined as: ;in: : The penalty value, the larger the value, the closer to the boundary and the higher the risk; : Time step weight coefficient, reflecting the difference in the impact of different time periods on life. The weight is determined based on the battery load sensitivity analysis and meets the normalization conditions ; : Penalty intensity coefficient, which controls the weight of the penalty term in the objective function and is adjusted through cross-validation; : A small constant to prevent division by zero. This penalty function is inversely proportional to the distance to the boundary. The smaller the distance, the greater the penalty, effectively preventing the strategy from getting too close to the safety envelope. By adding a life loss penalty term to the objective function, the objective function of the multi-objective optimization is adjusted to: ;in: is the original multi-objective weighted function of economy, life and grid constraints; is the boundary distance penalty term, which is an approximate expression of the life loss risk. Before outputting the final strategy, the envelope compliance verification is performed, and the boundary compliance test is performed on the final optimization strategy power sequence p* to ensure: ; And for the points near the boundary, a quadratic smoothing algorithm is used to adjust the power trajectory to ensure the continuity of the power change rate and prevent the impact of rapid jumps on the battery and power grid. The smoothing process meets the following requirements: ;Conditions and restrictions: ;in, It represents the sum of squares of differences in the power sequence at adjacent moments, and is used to measure the smoothness of power changes. A smaller value indicates a smoother change. K: The total number of time steps, which represents the number of discrete divisions in the entire optimization period. : The charge and discharge power value at the final optimization time k, which represents the actual dispatch power output of the energy storage system at that time; : The upper limit of the power safety envelope corresponding to time k, derived from the dynamic calculation of the life management module, represents the maximum safe allowable charge and discharge power at that moment. Key parameter determination, distance threshold δ: set according to the fatigue characteristics of the battery material and historical failure data, usually within 5% to 10% of the power boundary; weight coefficient : Through the analysis of experimental data on battery cycle aging sensitivity in different time periods, combined with the life model fitting, it is determined to reflect the different effects of load changes on life; penalty intensity coefficient γ: through multiple rounds of optimization simulation adjustment, the penalty term has a moderate impact on the optimization results to avoid over-constraint or under-constraint.

[0026] Specifically, the specific process of dynamically constructing the electricity price rule association network through transfer learning and generating confidence interval prediction curves for photovoltaic and load by combining quantile regression forest is as follows: extracting electricity price formation mechanism data from the electricity market historical database, parsing the correlation characteristics between different electricity price mechanisms through pre-trained prediction models, and constructing an electricity price rule knowledge framework; based on the electricity price structure characteristics of the target area, adapting the localized rule network through graph structure transfer learning, integrating multi-source operation data, and synchronously training the prediction model at multiple confidence levels through the quantile regression forest algorithm to generate probability distribution curves for photovoltaic output and load.

[0027] In this implementation, the electricity price rule association network is constructed. First, data related to the electricity price formation mechanism is extracted from the electricity market historical database, including electricity price adjustment periods, peak and valley distribution, and market trading rules. The pre-trained model is used to analyze the inherent relationships between electricity price mechanisms and construct an electricity price rule knowledge framework. This framework supports the application of subsequent transfer learning by representing the mapping relationship between different electricity price rules. Transfer learning adapts to the local electricity price rule network, and the target area electricity price structure is represented as a graph structure. , where the node set Indicates different electricity price periods, edge set Represents the regular association between time periods. Transfer learning aims to optimize the node weight vector To adapt to local electricity price characteristics and meet: ;in, is the transfer learning loss function, which measures the performance of the adjusted model on the local dataset. The fitting error of Represents the weight corresponding to the node in the electricity price rule network, reflecting the importance of each electricity price period in the local mechanism. The weight vector is iteratively updated by the gradient descent method. The initial value is provided by the pre-trained model. The update steps are: ; is the learning rate, the value of which is determined based on the performance of the validation set; The quantile regression forest generates a confidence interval prediction curve and trains the quantile regression forest model by integrating the localized electricity price rule network with multi-source operation data (such as photovoltaic output, load and meteorological data). Suppose the training sample is ,in: For the Sample feature vectors, including historical photovoltaic and load data; is the corresponding observation value. For the confidence level , the model outputs the predicted quantiles: ;in, is the number of trees in the forest, ensuring the generalization and stability of the model; For the A regression tree in time Confidence level The objective function of quantile regression is: ; The quantile regression loss function is defined as: ; is the residual; Is an indicator function that takes 1 when the residual is negative and 0 otherwise. This loss function ensures that the model reasonably penalizes the deviation of the predicted distribution at different confidence levels, achieving accurate estimation of the confidence interval. : According to actual business needs and historical data distribution, multiple confidence levels are selected, such as low (0.1), medium (0.5, representing the median), and high (0.9), to capture forecast uncertainty. This selection is based on empirical analysis of the balance between confidence interval width and coverage. Learning rate : Determine by adjusting the parameters of the validation set to ensure that the transfer learning process converges and avoids overfitting. Usually the initial value is set to 0.01 or 0.001, and dynamically adjusted according to the loss reduction curve. : Determined through cross-validation to ensure a trade-off between model complexity and generalization performance, usually an integer between 50 and 200 is selected.

[0028] Specifically, the specific process of extracting the net load change rate abrupt change feature sequence based on the dynamic time warping algorithm is as follows: net load mutation events are screened from historical data, standardized waveforms are extracted as abrupt change reference templates, the dynamic time warping distance between the current curve and the template is calculated in real time, and the candidate feature area is marked when the distance is lower than the similarity threshold; the change rate derivative is calculated within the feature area, and if the absolute value of the change rate of multiple consecutive sampling points exceeds the historical fluctuation range, it is determined to be a valid abrupt change feature, and the timestamp, duration, and maximum change rate of the feature area are encapsulated as a feature vector sequence; wherein abrupt changes include steep rises and steep falls.

[0029] In this implementation, the extraction of the abrupt change reference template involves first screening typical net load mutation events from historical net load data and extracting the corresponding standardized waveform as the abrupt change reference template. The standardization process includes normalizing the waveform amplitude to eliminate the impact of amplitude differences between different events on matching. Real-time dynamic time warping distance calculation and feature area identification, in real-time monitoring, uses the sliding window method to represent the current net load change curve segment as a sequence , the reference template is represented as a sequence ,in: is the number of sampling points in the current monitoring window; is the number of sampling points in the template sequence. The dynamic time warping distance is defined as: ;in, Represents the set of all allowed paths, used to align sampling points in the sequence; is the absolute distance between sequence points; For the path, ensure the nonlinear alignment of the time series. When the dynamic time warping distance Less than the set similarity threshold When , the current curve segment in the sliding window is marked as a candidate feature area. Parameter Description: Similarity Threshold: The threshold is determined by the statistical distribution of the distance of typical abrupt change events in historical data. Usually, the upper limit covering 95% of similar events is selected to ensure a balance between matching sensitivity and false alarm rate. Change rate derivative and abrupt change feature determination, in the candidate feature area, calculate the derivative sequence of the net load change rate: ;in, Indicates the The rate of change of each sampling point; is the sampling time interval. Based on the historical net load fluctuation data, the normal fluctuation range of the change rate is calculated. When continuous The absolute value of the rate of change of the sampling point All of them exceed the historical fluctuation range, which means: ; The area is judged to be a valid steep change feature, including steep rise (positive rate of change) or steep drop (negative rate of change). Parameter Description: To determine the number of continuous sampling points, it is usually determined empirically based on the sampling frequency and the duration of the abrupt change event; The maximum threshold of historical fluctuation is obtained through statistical analysis of historical net load change rate samples, for example, the 99% quantile is selected. Feature vector sequence encapsulation: for the feature area determined to be an effective steep change, the following parameters are extracted to form a feature vector sequence : ;in, The starting timestamp of the feature area; is the duration of the characteristic area; The maximum absolute value of the change rate in the characteristic area. This sequence provides dynamic input information of the net load mutation for the subsequent scheduling module. Key parameter determination description Similarity threshold: By statistically analyzing the DTW distance distribution of historical typical sudden change events, the upper limit of the 95% confidence interval is taken as the threshold to ensure detection sensitivity and control the false alarm rate. Continuous judgment points :Combining the sampling period and the actual duration of the abrupt change event, the empirical setting is 2 to 5 sampling points. Historical fluctuation threshold :Based on the historical net load change rate samples, the 99th percentile is taken as the standard for judging whether it exceeds the normal fluctuation range.

[0030] Specifically, when it is detected that the net load change rate in a continuous period exceeds the set threshold and the contract demand margin is lower than the safety margin, the specific process of dynamically compressing the upper limit of the battery discharge depth and activating load hierarchical regulation is as follows: Based on the steep change feature sequence output by the feature extraction module, identify that the net load change rate at multiple consecutive sampling points exceeds the upper limit of the historical normal fluctuation range. Synchronously calculate the consumed demand within the contract period through real-time integration, and deduce the occupancy trend of the remaining demand in the remaining period in combination with the prediction curve. When the margin is lower than the safety margin and there are steep change features, trigger the demand risk state; compress the upper limit of the discharge depth in stages according to the risk duration. In the initial stage, reduce a fixed proportion of the basic discharge depth, and add a compression amplitude for each newly added high-risk period. Synchronously activate the load priority management mechanism, immediately cut off the preset non-critical load categories, and delay the enabling time window of adjustable loads to reduce the peak pressure of the power grid through load-side coordination.

[0031] In this implementation plan, for the steep change feature recognition and contract demand margin judgment, using the steep change feature sequence output by the feature extraction module, identify that the net load change rates {q1, q2,... } of M consecutive sampling points all exceed the upper limit threshold δq of the historical normal fluctuation range, that is, satisfy: Parameter description: qm: The net load change rate at the m-th sampling point; δq: The upper limit threshold of the net load change rate, determined by historical fluctuation statistical data, generally selecting the 99% quantile to ensure accurate identification of abnormal steep changes; M: The number of consecutive sampling points, set according to the sampling frequency and system response requirements, usually 3 - 5 points. The consumed demand Dc within the contract period is calculated by real-time integration: ; where p(t): The real-time load power; t0: The start time of the contract period; t: The current time. Predict the occupancy trend of the remaining demand Estimated through the prediction curve ŝ(t): ; where ŝ(τ): The predicted load power in the future period; te: The end time of the contract period. The contract demand margin R is calculated as: ; where : The maximum demand limit specified in the contract. When R < Rsafe (safety margin threshold) and there are steep change features, trigger the demand risk state. Parameter description: Rsafe: The safety margin threshold, determined by analyzing historical demand overrun risks and system tolerance, generally taking 5% - 10% of the contract demand limit. Dynamic compression strategy for the upper limit of the discharge depth, according to the demand risk duration , dynamically adjust the upper limit of the battery discharge depth : ; where : The upper limit of the basic discharge depth; : The cumulative number of risk periods; : The The compression ratio corresponding to each risk period satisfies , and increase. The specific setting is: initial risk period compression ratio (Fixed ratio), the compression ratio increases with each new risk period: ;in, : Initial compression ratio, determined by experience and safety analysis, usually ; : Incremental step size to ensure that the compression force is gradually increased as the risk accumulates. Synchronously activate the load priority management mechanism and implement the following measures: Cut off non-critical loads: Immediately disconnect the preset non-critical load categories to reduce the total load. Delay the adjustable load activation time window: Adjust the operating time of the adjustable load, postpone the startup, and alleviate the peak pressure. This mechanism is based on a predefined load level and priority system to achieve hierarchical management of different categories of loads and ensure the normal operation of critical loads. Net load change rate threshold :Through statistical analysis of historical load change rate data, select Quantiles are used as thresholds to ensure accurate detection of abnormal changes. Safety margin threshold :Based on the risk assessment of demand exceeding the limit and the contract agreement, it is set as the maximum demand of the contract , ensuring a reasonable risk tolerance. Compression ratio and incremental step size :Determined based on battery safety specifications and actual scheduling needs, initial compression ensures a certain buffer, and the incremental step size reflects the risk accumulation effect. and the cumulative number of risk periods :According to the sampling frequency and system response speed, ensure timely response and cumulative identification of risk status.

[0032] Specifically, within the constraints of the safety envelope, a multi-objective evolutionary algorithm guided by tabu search is used to simultaneously optimize the economic objectives, life loss objectives, and grid constraint objectives to generate a charging and discharging strategy with a three-objective balance. The specific process is as follows: the power safety envelope is converted into a time-varying inequality constraint set to define the boundary of the feasible domain of charging and discharging power; the initial solution set is generated by guiding the tabu search algorithm to ensure that the candidate strategies are distributed within the feasible domain, and the three objective function values ​​are simultaneously evaluated for each candidate scheme. The economic objective calculates the sum of time-of-use electricity cost and demand penalty, the life loss objective quantifies the cumulative effect of charge and discharge rate stress, and the grid constraint objective evaluates peak power and volatility penalty, and the Pareto optimal solution set is screened through non-dominated sorting; an adaptive penalty term inversely proportional to the boundary distance is injected into the solution approaching the safety envelope boundary, and a taboo list is activated for the out-of-bounds solution to block repeated generation. The output charging and discharging strategy is a structured instruction sequence, including: charging and discharging power values ​​corresponding to the timestamp sequence, and a load grade adjustment action timing list.

[0033] In this implementation, the power safety envelope is converted into a time-varying inequality constraint, and the safety envelope is expressed as a power constraint interval on the time series t=1,2,…,T: ; Parameter description: : charging and discharging power at time t; : the lower boundary power limit of the safety envelope at time t; : upper boundary power limit of the safety envelope at time t; : The total number of time steps in the scheduling cycle. This constraint defines the feasible region boundary of the charging and discharging power, ensuring the safe operation of the strategy. The initial solution set is generated by the tabu search algorithm (TS) within the constraint domain. , each solution satisfy: ; Tabu search avoids local optimal traps through neighborhood search and taboo list, promoting solution diversity and global exploration. Multi-objective function simultaneous evaluation, for each candidate strategy Simultaneously calculate three objective functions: economic objective The sum of time-of-use electricity cost and demand penalty: ; Parameter description: : electricity price at time t; : forward discharge power; : Demand penalty weight coefficient, set according to the contract terms; : Strategy maximum demand; :The maximum allowable demand in the contract. Life loss target Quantifying the cumulative effect of charge and discharge rate stress: ; Parameter description: : Time weight, reflecting the battery aging sensitivity in different time periods; :Battery rated capacity; : Ratio stress index, usually set to 2~3 to reflect the nonlinear stress effect. Evaluate peak power and power fluctuation penalty: ; Parameter description: : Peak power penalty weight coefficient; : Power fluctuation penalty weight coefficient. Pareto optimal solution screening and penalty mechanism, perform non-dominated sorting on the initial solution set, screen out the Pareto optimal solution set, and ensure multi-objective balance. For solutions close to the safety envelope boundary, calculate the minimum distance to the boundary: ; and inject an adaptive penalty term into the objective function: ; Parameter description: : Penalty weight coefficient, determines the penalty intensity; : A small constant to prevent division by zero. Immediately add the out-of-bounds solution to the taboo list, prohibit repeated generation, and ensure the legality of the solution set. Output structured charge and discharge strategy, the final output strategy includes: timestamp sequence corresponding to the charge and discharge power value ;Load classification adjustment action sequence list, record load adjustment category and execution time. Determination method description weight coefficient The system is determined through historical data fitting and system safety assessment, taking into account the balance between economy and safety; the safety envelope boundary is dynamically generated by the life management module to ensure real-time risk constraints; the non-dominated sorting adopts the classic NSGAII algorithm to ensure multi-objective optimization effects; the taboo list length and neighborhood structure design are adjusted based on the problem scale and computing resources to ensure search efficiency and globality.

[0034] Specifically, when the distance falls below the warning threshold, rolling optimization control is initiated, and the life protection weight is increased according to a preset ratio and the economic constraints are relaxed. The specific process is as follows: the minimum Euclidean distance from the actual power point to the envelope boundary is continuously calculated, and the rolling optimization mechanism is triggered when the distance value falls into the preset warning interval; the multi-objective function is reconstructed within the rolling time window, the life protection target weight is increased to a fixed multiple of the baseline value, and the allowable deviation range of the economic constraints is simultaneously relaxed. The power derating command is dynamically generated through the quadratic programming solver, and the derating amplitude is negatively exponentially related to the distance value to ensure that the power trajectory converges safely along the envelope.

[0035] In this implementation, the minimum Euclidean distance from the power point to the envelope boundary is calculated in real time. Assuming the current power point is q, the corresponding upper and lower boundaries of the envelope are L (lower boundary) and U (upper boundary), and the calculated distance d is: ; Parameter description: : Actual charging and discharging power at the current moment; : The lower limit power of the safety envelope at the current moment; : The upper boundary power of the safety envelope line at the current moment; : The minimum Euclidean distance from the actual power point to the envelope boundary. Entering the preset warning zone When the rolling optimization control mechanism is triggered. Warning interval determination method: set according to battery safety operation experience and system dynamic response characteristics; combine historical operation data and life attenuation risk threshold to set a reasonable range to ensure timely warning and avoid false triggering. Rolling optimization mechanism and time window definition, set the rolling optimization time window length to ,dynamic multi-objective optimization is performed within this window, and the control strategy is continuously updated to ensure real-time response.,Multi-objective function reconstruction and weight adjustment,The multi-objective function is reconstructed as: ; Parameter description: : Life protection objective function, reflecting battery life loss; : Economic objective function, reflecting electricity cost and demand penalty; : Grid constraint objective function, reflecting peak power and fluctuation rate; , , : Objective function weight coefficient. When rolling optimization is triggered, the weight coefficient is adjusted to: ; Parameter description: , , : Baseline weight before rolling optimization; : Life protection weight enhancement multiple, usually set according to battery safety requirements; The economic constraint allows for deviations, reflecting the degree of relaxation, and is set based on risk tolerance. This adjustment prioritizes increasing the weight of life protection, relaxing economic constraints, and balancing safety and economy. The dynamic generation of power derating instructions uses a quadratic programming solver within a rolling window: ; Constraints: ;in: : Power decision variables for the next W moments; : positive definite weight matrix, reflecting the quadratic term of the weighted objective function, including the life protection impact after weight adjustment; : linear term vector, including electricity price and demand penalty factors; : Linear constraint matrix and vector of the system and power grid. and distance The negative exponential relationship is defined as: ; Parameter description: : Power derating ratio; : Maximum derating, based on battery and grid load limits; : Distance sensitivity coefficient, which controls the rate of change of derating with distance and is adjusted according to the system response performance; : The minimum Euclidean distance from the power point to the envelope boundary. This formula ensures that the smaller the distance (i.e., the closer to the envelope boundary), the greater the derating, achieving continuous convergence of the safe power trajectory. Result feedback and closed-loop control: The power derating command is sent to the energy storage EMS execution unit in real time, and the adjusted power trajectory is fed back to the life management module and the envelope dynamic generation unit to achieve closed-loop risk control. Parameter determination instructions, warning interval : Based on the battery chemical characteristics and the risk of excessive lithium deposition, combined with statistical analysis and simulation verification, to ensure sensitivity without excessive triggering; life protection weight is increased multiples : Adjusted according to the battery life attenuation rate and system safety requirements, generally 1.5~3; economic constraint relaxation range : Make appropriate adjustments based on economic operation needs and risk assessment to avoid excessive impact on costs; the maximum reduction range Distance sensitivity coefficient : Obtained through historical operation data fitting and on-site debugging to ensure response speed and safety margin.

[0036] Specifically, the specific process of feeding back the actual lithium plating risk characteristics to the life management module to drive the online update of the safety envelope is as follows: real-time collection of charge and discharge current waveforms, temperature distribution and voltage fluctuation data, reconstructing the internal ion concentration gradient field of the battery based on impedance spectrum analysis, locating the abnormal diffusion delay area and quantifying the local ion flux deviation; inputting the measured ion flux deviation into the electrochemical parameter estimation algorithm, updating the mobility and diffusion coefficient in the ion transport equation, recalculating the lithium plating risk level according to the updated ion transport equation, and generating a power safety boundary curve that continuously evolves over time; injecting the new envelope parameters into the constraint library of the hierarchical optimization module in real time through the dynamic interface, replacing the original boundary value and triggering strategy re-verification.

[0037] In this implementation, the real-time data acquisition and ion concentration gradient field reconstruction system collects key operating data of the energy storage battery in real time, including: charge and discharge current waveforms ;Battery temperature distribution ;Voltage fluctuation signal Based on electrochemical impedance spectroscopy (EIS) analysis and combined with the internal physical model of the battery, the ion concentration gradient field of the positive and negative electrodes of the battery is reconstructed. ,in: ; Parameter description: :Space-time coordinates The ion concentration at : Equilibrium ion concentration reference value; : Dynamic deviation of concentration over time and space. By identifying abnormal areas of diffusion delay (Collection of internal regions of the battery) to locate local ion flux anomalies. Quantify the local ion flux deviation and input electrochemical parameter estimation, and define the local ion flux deviation as: ; Parameter description: : measured ion flux; : Reference base flux (based on steady state or historical mean). Input electrochemical parameter estimation algorithm for ion mobility and diffusion coefficient Updated estimate: ; ; Parameter description: : previously estimated mobility and diffusion coefficients; :Updated parameters; : the temporal and spatial average of the local ion flux deviation within the region; : Weight coefficient, used to adjust the parameter update rate. The public method is based on historical data regression fitting to prevent over-adjustment and ensure model stability. The lithium plating risk level is recalculated based on the ion transport equation with updated parameters, using the classic ion transport partial differential equation model: ; Parameter description: : ion concentration distribution function; : electric field vector, representing the potential gradient driving migration; , : The diffusion coefficient and mobility updated in the above steps. Based on the newly calculated concentration gradient and ion flux distribution, the lithium plating risk level is dynamically calculated. : ; Parameter description: :Real-time lithium plating risk level; : ion concentration gradient amplitude; :Lithium deposition critical concentration gradient threshold is pre-set based on the physical properties of battery materials. Power safety boundary curve is generated and dynamically updated according to the risk level. Dynamically adjust the charge and discharge power safety envelope: ; Parameter description: :time The safe power boundary value; : Initial safety margin power value; : Risk sensitivity coefficient, reflecting the degree of influence of risk level on power limit. The boundary curve evolves continuously over time, reflecting the actual operation risk status. Real-time injection of envelope parameters and strategy re-verification are implemented through the dynamic interface module to update the updated envelope parameters in real time. Inject the constraint library of the hierarchical optimization module and replace the original boundary conditions: Replace constraints: ;in, The decision variable for charge and discharge power in the current optimization strategy. Triggering the re-verification of the hierarchical optimization module strategy to ensure that the strategy still meets the optimization objectives and safety constraints under the new boundary. Parameter determination method, weight coefficient : Based on historical measured data and battery life degradation test results, the critical concentration gradient is determined by least squares fitting or Bayesian estimation method, taking into account the model response sensitivity and stability. Provided by the battery manufacturer or obtained through laboratory lithium precipitation threshold testing, reflecting the electrochemical limit of the material; risk sensitivity coefficient : Based on safe operation experience and risk management strategy settings, ensure that power limits are reasonable and avoid excessive conservatism or risk omissions.

[0038] In summary, this application has at least the following effects: This multi-objective optimization-based industrial and commercial energy storage EMS dispatching system uses real-time analysis of internal battery ion migration states and dynamic generation of charge and discharge power safety envelopes to effectively warn and mitigate lithium plating risks, prevent internal battery structural damage and performance degradation, and extend battery life. Transfer learning is used to construct a price-rule association network, combined with quantile regression forests to generate photovoltaic and load confidence interval forecasts. Dynamically extracting characteristics of sudden net load changes enables accurate identification and response to grid load fluctuations, improving the dispatching solution's adaptability to complex industrial and commercial environments. A multi-objective evolutionary algorithm guided by tabu search optimizes charging and discharging strategies while meeting the hard constraints of the safety envelope, taking into account economic efficiency, life loss, and grid constraints. This optimizes the overall operational efficiency of the energy storage system. The execution arbitration module monitors the distance between power and the safety margin in real time, and uses a rolling optimization control strategy to dynamically adjust life protection weights and economic constraints to achieve safe convergence of the power trajectory. Real-time lithium plating risk characteristics are fed back to the life management module to drive online updates of the safety envelope, ensuring timely and accurate system responses. By dynamically compressing the discharge depth and activating the load grading adjustment mechanism, the peak pressure on the power grid can be effectively reduced, the load structure can be optimized, and the efficient and coordinated operation of industrial and commercial energy storage systems in complex power grid environments can be supported.

[0039] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0040] The present invention is described with reference to flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts 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, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0041] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0042] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0043] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0044] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization is characterized by: It includes the following modules: life management module, feature extraction module, layer optimization module, and execution arbitration module; The life management module is used to analyze the internal ion migration state of the battery in real time, calculate the lithium plating risk coefficient based on the ion concentration gradient distribution, and dynamically generate the charge and discharge power safety envelope, which is embedded in the hierarchical optimization module as a hard constraint condition; The feature extraction module is used to dynamically construct an electricity price rule association network through transfer learning, generate confidence interval prediction curves for photovoltaic and load by combining quantile regression forest, and extract the characteristic sequence of sudden change of net load change rate based on dynamic time warping algorithm; The hierarchical optimization module is used to dynamically compress the upper limit of the battery discharge depth and activate load hierarchical regulation when it detects that the net load change rate in consecutive time periods exceeds a set threshold and the contract demand margin is lower than the safety margin. Within the constraints of the safety envelope, the multi-objective evolutionary algorithm guided by tabu search simultaneously optimizes the economic goal, life loss goal and grid constraint goal to generate a charging and discharging strategy that balances the three goals; The execution arbitration module is used to monitor the distance between the charge and discharge power and the safety envelope in real time. When the distance is lower than the warning threshold, rolling optimization control is initiated, the life protection weight is increased according to a preset ratio, and the economic constraints are relaxed. At the same time, the actual lithium plating risk characteristics are fed back to the life management module to drive the online update of the safety envelope.

2. The industrial and commercial energy storage EMS scheduling system based on multi-objective optimization according to claim 1 is characterized by: The specific process of real-time analysis of the internal ion migration state of the battery, calculation of the lithium plating risk coefficient based on the ion concentration gradient distribution, and dynamic generation of the charge and discharge power safety envelope is as follows: Real-time collection of internal battery sensor data streams, generation of ion concentration gradient distribution maps in the positive electrode region through ion transport equations, and dynamic calculation of local ion flux benchmark values; Analyze the concentration mutation nodes in the gradient distribution diagram and calculate the real-time deviation between the measured ion flux in this area and the baseline value; The deviation is mapped to the lithium plating risk coefficient. When the risk coefficient exceeds the threshold, the upper limit of the charge and discharge power is adjusted according to the risk growth rate to generate a power safety envelope that evolves continuously over time. The slope of the envelope boundary is positively correlated with the risk coefficient increment.

3. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization according to claim 2 is characterized by: The specific process of embedding the envelope into the hierarchical optimization module as a hard constraint is as follows: During the initialization phase of the hierarchical optimization module, the safety envelope parameters are loaded and the charge and discharge power constraint matrix is ​​constructed. During the optimization iteration, the power trajectory of the candidate strategy is detected in real time. If the trajectory exceeds the envelope boundary, it is marked as an infeasible solution through the tabu search mechanism. An adaptive penalty function is activated for solutions close to the boundary, and a life loss penalty term inversely proportional to the boundary distance is added to the objective function; Envelope compliance verification is performed before outputting the final strategy, and secondary smoothing is performed on the boundary point strategy to ensure the continuity of the power change rate.

4. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization according to claim 1 is characterized by: The specific process of dynamically building an electricity price rule association network through transfer learning and generating confidence interval prediction curves for photovoltaic and load by combining quantile regression forest is as follows: Extracting electricity price formation mechanism data from the electricity market historical database, analyzing the correlation characteristics between different electricity price mechanisms through pre-trained prediction models, and building an electricity price rule knowledge framework; Based on the electricity price structure characteristics of the target area, the localized rule network is adapted through graph structure transfer learning, multi-source operation data is integrated, and the prediction model is simultaneously trained at multiple confidence levels through the quantile regression forest algorithm to generate probability distribution curves of photovoltaic output and load.

5. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization according to claim 4 is characterized by: The specific process of extracting the characteristic sequence of sudden change of net load change rate based on the dynamic time warping algorithm is as follows: Filter net load mutation events from historical data, extract standardized waveforms as abrupt change reference templates, calculate the dynamic time warping distance between the current curve and the template in real time, and mark candidate feature areas when the distance is below the similarity threshold; The derivative of the rate of change is calculated within the feature area. If the absolute value of the rate of change of multiple consecutive sampling points exceeds the historical fluctuation range, it is determined to be a valid steep change feature, and the timestamp, duration, and maximum rate of change of the feature area are encapsulated into a feature vector sequence. The steep changes include steep rise and steep fall.

6. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization according to claim 5 is characterized by: When it is detected that the net load change rate in consecutive time periods exceeds the set threshold and the contract demand margin is lower than the safety margin, the specific process of dynamically compressing the battery discharge depth upper limit and activating load graded regulation is as follows: Based on the abrupt change feature sequence output by the feature extraction module, it identifies when the net load change rate at multiple consecutive sampling points exceeds the upper limit of the historical normal fluctuation range. It then simultaneously calculates the consumed demand within the contract period through real-time integration. It then uses the forecast curve to deduce the demand occupancy trend for the remaining period. When the remaining capacity falls below the safety margin and abrupt change features are present, a demand risk state is triggered. The upper limit of the discharge depth is compressed in stages according to the duration of the risk. In the initial stage, the fixed proportion of the basic discharge depth is reduced. The compression amplitude is increased with each new high-risk period. The load priority management mechanism is activated simultaneously, the preset non-critical load categories are immediately cut off, and the activation time window of the adjustable load is delayed. The peak pressure of the power grid is reduced through load-side collaboration.

7. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization according to claim 6 is characterized by: Within the constraints of the safety envelope, a multi-objective evolutionary algorithm guided by tabu search is used to simultaneously optimize the economic objective, life loss objective, and grid constraint objective. The specific process of generating a charge-discharge strategy that balances the three objectives is as follows: Convert the power safety envelope into a time-varying inequality constraint set to define the boundary of the feasible region of charging and discharging power; The tabu search algorithm guides the generation of the initial solution set, ensuring that the candidate strategies are distributed within the feasible region. Three objective function values ​​are simultaneously evaluated for each candidate solution: the economic objective calculates the sum of time-of-use electricity costs and demand penalties, the life loss objective quantifies the cumulative effect of charge and discharge rate stress, and the grid constraint objective evaluates peak power and volatility penalties. Finally, a non-dominated sort is used to screen the Pareto optimal solution set. An adaptive penalty term inversely proportional to the distance to the boundary is injected into the solution approaching the boundary of the safety envelope. A taboo list is activated to block repeated generation of out-of-bounds solutions. The output charging and discharging strategy is a structured instruction sequence, including: the charging and discharging power values ​​corresponding to the timestamp sequence, and the load classification adjustment action timing list.

8. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization according to claim 1 is characterized by: When the distance falls below the warning threshold, rolling optimization control is initiated. The specific process of increasing the life protection weight according to the preset ratio and relaxing the economic constraints is as follows: Continuously calculate the minimum Euclidean distance from the actual power point to the envelope boundary, and trigger the rolling optimization mechanism when the distance value falls into the preset warning range; The multi-objective function is reconstructed within a rolling time window, and the weight of the lifetime protection objective is increased to a fixed multiple of the baseline value. The allowable deviation range of the economic constraint is simultaneously relaxed. The power derating command is dynamically generated through the quadratic programming solver. The derating amplitude has a negative exponential relationship with the distance value, ensuring that the power trajectory converges safely along the envelope.

9. The industrial and commercial energy storage EMS dispatching system based on multi-objective optimization according to claim 8 is characterized by: At the same time, the actual lithium plating risk characteristics are fed back to the life management module to drive the online update of the safety envelope. The specific process is as follows: Real-time collection of charge and discharge current waveforms, temperature distribution, and voltage fluctuation data. Based on impedance spectroscopy analysis, the battery's internal ion concentration gradient field is reconstructed to locate areas of abnormal diffusion delay and quantify local ion flux deviations. The measured ion flux deviation is input into the electrochemical parameter estimation algorithm to update the mobility and diffusion coefficient in the ion transport equation. The lithium plating risk level is recalculated based on the updated ion transport equation to generate a power safety margin curve that continuously evolves over time. The new envelope parameters are injected into the constraint library of the hierarchical optimization module in real time through a dynamic interface, replacing the original boundary values ​​and triggering strategy re-verification.

Images