AGORA FAQ
- General Questions
- Functionalities & Benefits
- System Configuration
- SCADA/EMS & Hardware
- Methodologies
What is AGORA?
What other products does it compare to?
Could you briefly enumerate the functionalities of AGORA?
What are the main selling points of AGORA?
What are the main selling points of the AGORA Restoration functionality?
What are the technical limitations on network size?
What are the input information requirements?
How do I incorporate operational best practices?
What are the unique features of AGORA?
Does AGORA display the results graphically?
What is AGORA?
AGORA is a suite of intelligent software systems, which complements an existing SCADA and EMS system.We can think of it as a “next-generation” EMS, since it provides a number of higher-level functionalities, which are not present in any other EMS vendor. It is the only system in the world capable of producing real-time action plans in terms of concrete controlling maneuvers, thanks to its own highly reliable Load-Flow solver and State Estimator computing all the way up to voltage collapse. It is therefore an invaluable decision support tool for the operator in three key stages:
- Normal Conditions: State Estimation.
- Alert Conditions: Generation of Corrective actions plan under Violation of Operational Limits.
- Alarm or Emergency Conditions: Generation of a Blackout or Disturbance Restoration Plan
It is important to remark that the system is based on the full physical grid model, and that the proprietary HELM™ methodology for power flow computations is the key basic block on which all other algorithms rely. This is why all control actions are fully safe (from the steady-state point of view point), as they are electrically verified at all times.
Other side benefits derived from its innovative HELM™ power flow and State Estimation techniques are more reliable real-time Contingency Analysis, PV/QV Analysis Curves, and general “on-line” simulations. HELM always provides correct results in an unattended fashion, since the method is constructive and direct (i.e. it is not iterative, such as Newton-Raphson or similar methods, and therefore its results do not depend on choosing “the right” starting point).
What other products does it compare to?
AGORA compares to SCADA/EMS systems for managing electrical grids among which we can mention GE, SIEMENS, ABB and AREVA systems, but AGORA was in fact designed to complement this types of SCADA/EMS systems bringing its unique functions, not found in other products, to the operation and planning personnel. Its deployment is non-intrusive and thus can be easily integrated to existing SCADA/EMS.
Could you briefly enumerate the functionalities of AGORA?
In the following we list AGORA’s main modules that can be divided into two groups: the key differentiating modules not present in other EMS tools, and common modules present in other EMS tools that differ from the standard products because they rely on AGORA’s unique Load flow methodology.
Key differentiating modules:
- State Estimator
- Limit violation Solver
- Restoration
Common modules:
- Real time simulator
- PV-QV curves
- Contingency analysis
- Customization for planned outages analysis
What are the main selling points of AGORA?
We list here many key benefits of using AGORA system:
- Our State Estimator (SENSE) provides consistent, reliable results virtually 100% of the time, even with low levels of RTU measurement visibility. We use our power flow calculation to confirm the results of every state estimation by introducing the power flow equations as restrictions in the least square minimization process.
To put it another way, the AGORA state estimator will never provide a result that is not a physically-realizable grid condition. Of course, this means more accurate and more reliable state estimation results, providing greater reliability and security for transmission system operations.
There is no maintenance or sophisticated tuning required of the State Estimator; only some configuration. The resulting base cases can be utilized with the assurance that they are a valid system-wide representation of the conditions present in the grid at that moment in time.
- All of the power flow calculations are reliable and accurate. Because of the robustness of our algorithm, we can provide accurate real-time PV and QV curve generation, perform reliable Contingency Analysis, and support Simulation in Real-Time. We also provide accurate Fault Analysis (three-phase and unsymmetrical faults) results.
- Our load flow calculation methodology enables us to provide two real-time tools which are truly unique in the market:
- Limits Violations Solver: This tool provides the operator with an optimized plan of actions in order to correct stressful conditions on the grid. These plans may contain any of the actions that are actually available to the operator through the SCADA, and they are always fully checked on the real-time base case through load-flow computations.
- Blackout Restoration Solver: This tool provides the operator with a restoration plan for outages of any size, including large scale blackouts. As above, the system always checks the validity of the plan on the current real-time condition through load-flow computation. Accordingly, if the state of the grid (as seen through the state estimation) changes so much that the current plan becomes unfeasible, the system re-computes a new plan.
Thanks to these effective real-time tools, AGORA becomes the most advanced and effective system for maintaining reliable system operation at all times, allowing operators to maximize the availability of electric power for all customers.
AGORA has been installed in several networks, ranging from a sub-transmission and distribution network for the largest distribution company in Spain, to a hierarchical, nationwide installation for the transmission and sub-transmission Mexican system.
AGORA is able to manage the multiple complexities found in the different networks because of its approach based strictly in action simulation thought calculation, making it ideal for the new challenges that the Smart Grid will pose to all the new decision-tool applications.
What are the main selling points of the AGORA Restoration functionality?
Bellow we list the most important AGORA’s Restoration benefits for Top Management, local Government and Public Utilities Commissioners.
- AGORA Intelligent Grid Monitoring and its Innovative technology helps in improving the overall score on quality indexes established by regulators (power losses, time of interruptions, etc.) having impact on BUDGET with a short period for ROI. Minimizing unscheduled interchanges and curbing overdraw improves Security. Providing Reliable knowledge of the grid state for operations or market purposes.
- Improvement of the Public Image of the industry in society and in the rest of the State and Country thanks to the above, and facilitating auditing of grid management practices through information provided.
- Increase in the availability of electricity supply state and nationwide, by means of the increase of reliability of Operation (transmission capacity, generation assets management, Forecasting demand,..) even nearer to voltage collapse. Optimize Planned optimal rolling blackouts when load shedding is necessary.
- Auditable processes all the way through system management with AGORA. AGORA Intelligent Grid Monitoring and its Innovative technology helps in improving the overall score on quality indexes established by regulators (power losses, time of interruptions, etc.) having impact on BUDGET with a short period for ROI. Minimizing unscheduled interchanges and curbing overdraw improves Security. Providing Reliable knowledge of the grid state for operations or market purposes.
What are the technical limitations on network size?
With the server hardware currently available (as of 2014), the concrete recommendations given in questions included guarantee a satisfactory and prompt performance of the system when the network comprises of up to about 4,000 effective electrical nodes. Under these conditions, for instance, the State Estimator typically finds the solution in about 30 seconds, even when the automatic topology estimation of some model parameters (switch & breaker status) is enabled.
But large network sizes increased by electrical nodes not having values read from the SCADA can be also treated with some limitations which depends on a case by case grid model, reaching experimental sizing in the 10.000 range for the number of nodes.
What are the input information requirements?
The data input that the AGORA system requires in order to have a successful integration with the existing SCADA or EMS can be separated in two big conceptual groups:
- Network model: model parameters of all the power system elements, together with the detailed bus-breaker- oriented topology. The required information for a network model is divided into three sets:
1. List of equipment elements and their connectivity
2. Electric attributes of the equipment
3. Measurement points definitions (sometimes known as “SCADA tags”)
- Telemetry data: i.e. the digital and analog measurements for voltages, flows, statuses, and any other changing magnitudes that are needed to characterize the electrical state of the power system. All of AGORA’s advanced analytical and problem-solving functionalities rely on the electrical states provided by AGORA’s own State Estimator. In turn, the Estimator requires various types of real-time measurements, both analog and digital. The Restoration and Limits-Violation solver modules may make use of a few other additional signals, most notably alarms for the automatic recognition of unavailability status of equipment for avoiding proposing their use in any manoeuvres in the plan generated.
How do I incorporate operational best practices?
By configuring the different modules (state estimation, limit violation solver, restoration, etc.) following the company operations standards and best practices.
Intensive testing is recommended using AGORA on simulation mode or better off AGORA connected with DTS or OTD simulating real time connectivity.
What are the unique features of AGORA?
AGORA contains two advanced network applications which no other vendor provides:
- Operations limit violation solver
- Automatic dynamic generation of black-out restoration plans
And AGORA provides unique features on other network applications:
- State estimator
- Real time simulation
- Contingency analysis
- PV-QV curves
The main functions of AGORA system compared with similar functionalities from other vendors are shown in the following table.
FUNCTIONALITY | OTHER VENDORS | AGORA |
Load Flow Algorithm |
|
|
State Estimator |
|
|
PV/QV Curves |
|
|
Contingency Analysis |
|
|
Limit Violation Solver |
|
|
Disturbance and Blackout Restoration |
|
|
Does AGORA display the results graphically?
Yes, AGORA dispose of numerous forms for displaying results (diagrams, charts, tables, etc.) that are user configurable, using a graphical user interface (GUI) implemented on top of the ILOG Views package that allows for system portability.
Displays are designed to be maintenance-free and automatically updated when the model is updated. The control and navigation of all system displays and actions required by the system operators can be performed using a mouse input device only.
A setup screen exists to define the operation of every required module. Each real-time graphic display includes all appropriate information available from the SCADA system, state estimator, or load flow calculation, in order to allow for rapid determination of system conditions. The graphical elements of each screen are customizable to match up with graphical elements of existing EMS/SCADA systems. The displays available include:
- One-line representation of the system
- Substation-level bus one-line diagrams
- Grid-level connectivity diagrams
- Bar graph displays of system voltages, phase angles, and power flows
The colors of buses and lines in real-time system screens are changed automatically to reflect states of normal operation, over limits, and under limits. All colors are user definable.
State Estimator
Can you briefly enumerate the features of the State Estimator module?
The AGORA State Estimator algorithm performs electrical coherence tests upon the field measures at each estimation cycle. The system makes uses of load profiles for points which lack measurements and is able to update the load profiles according to the estimation results using exponential smoothing. Some of its main features are:
- Active- Reactive decoupling
- Non probabilistic weights
- Artificial Intelligence technique using topology plausibility HELM™ load flow computation as a last check of electrical feasibility
Mainly used to detect model problems or others, as well as a robust state estimator near voltage collapse.
Can AGORA State Estimator do parameter estimation? And thus update its internal network model based on real-time information?
Yes, and the user can select which parameters should be analyzed by the system in order to have values suggested by AGORA state estimator after several estimations, but the model parameters changes that are suggested should be applied manually to the original model.
Can AGORA State Estimator do connectivity estimation? Can it do it at the detailed level of switches and breakers? And thus update its internal network model based on real-time information?
AGORA can estimate the status of circuit breakers and switches based on the real-time information received. The user can select which substations are going to be analyzed for the connectivity estimation or leave the system decide this.
Limit Violation Solver
Can you briefly enumerate the features and benefits of the Limit violation solver module?
This module has been designed to help the operator to solve operational limit violations in real-time. This functionality provides the user with the required actions to solve limit violations arising in the network and so it is preventive with regards to network events of alert and alarm. Its main features are:
- Detects limit violations on voltages and loads
- Under request proposes a sequence of actions to partially, progressively or totally solve the limit violations
- Heuristics guiding best practices of actions can be tuned with conformation priorities following utility best operational practices.It can save hundreds of thousand of dollars in minutes (client case)
This module is mainly used for solving issues in real time.
Restoration
Can you briefly enumerate the features and benefits of the Restoration module?
The Restoration module is a unique module not present in any other EMS tool in the market. It generates restoration plans in real time that are specific to the place and time of the disturbance using algorithms based on analytical modelling. The generation of these plans takes into consideration the electrical system priorities and limits and will not include any non-acceptable intermediate system conditions from a static stability view point. Some of its main features are:
- Proposes a sequence of actions: plan to restore the system to pre-disturbance state or close to it under certain criteria of optimality (time, load recovered, etc.).
- Rectifies to a new plan like a GPS when condition changes: actions performed by operators or expected effects.
- Optimal path algorithms and heuristics guiding best paths of actions
- Considers black start cranking paths to nuclear facilities among others
- Optimal islanding growth and connections
Mainly used for training, simulations and plans evaluations as well as for supporting operators at all times with a restorations plan at hand.
How does Restoration module handles dynamic stability?
This is defined in the rules configuration for Restoration. Rules-editor introduce decision rules for well known unstable situations leading to dynamic instabilities. When such a situation is encountered during Restoration, the plan is calculated avoiding paths leading to dynamic instabilities.
Future version in the roadmap of the Restoration module, will include testing for dynamic instabilities for group of actions additional to the rules defined.
P-V / Q-V curves
Does AGORA compute P-V and Q-V curves in real time?
The system provides a PV-QV computation function that computes PV-QV curves in real time. This is again possible thanks to the proprietary power flow algorithm (HELMTM power flow method) that is able to perform accurate power flow computations even near the voltage collapse.
The PV-QV curves module allows the user to define load variation zones and generation variation zones. These definitions can be combined to study the system behaviour when all the loads in a given zone increase keeping the P/Q rate constant and when all the generators in the same zone are compensating for the active power increase.
The user chooses the points where voltage will be measured and the system plots the PV-QV curves starting from the present state up to voltage collapse for these points as well as for any flow gate that the user may have defined.
PV-QV curves can also be computed taking into account the effect of a contingency that is answering when the voltage collapse will arise given that the chosen particular contingency happens. The results are plotted against the original PV-QV curve to show the change.
Contingency Analysis
How do I perform a Contingency Analysis?
The AGORA system provides a Contingency Analysis application that performs several types of analysis on user-defined zones of the network. The contingencies are computed using the full Load Flow calculation to ensure the reliability of the results. In AGORA no false warning of an inexistent contingency should happen, as well as it should never miss a real contingency. This is a consequence of the use of the fully reliable AGORA Load Flow algorithm: HELMTM method.
Besides the usual N-1 Contingencies as well as sets of those, AGORA can compute other types of contingency analysis:
- Lists of Sets of Actions (manoeuvres):the user may define lists of manoeuvres. This may represent the simultaneous loss of several equipment defining a complex contingency. The system is able to process lists of complex contingencies defined by the user.
- Contingencies on equipment that have their limits violated: the system may perform an N-1 contingency analysis on all the equipment on a user defined zone that have a limit violation. There are four possible choices: only for overloads, only for over voltages, only for under voltages or for any kind of limit violation.
- Pre-conditions: the lists of manoeuvres can also be used to define a set of initial actions on the power system to be carried out before performing the contingency analysis.
The contingency analysis to be performed automatically can be scheduled by the system at a given time on a batch (background) mode.
The analysis results are shown in detailed reports where the contingencies are ordered by severity order and where detailed information of the limit violations arising is provided.
Is AGORA capable of automating N-1 and N-1-1 contingency analysis?
AGORA can automate the standard N-1 contingency analysis, and also pre-configured lists of “complex” contingencies, which engineers in the industry may call N-k (where each contingency is actually a set of predefined actions, typically outaged elements).
In addition, AGORA Limit Violation Solver module may help identify the corrective actions that are required to fix the violations after the contingencies occur.
Model cleansing
How does AGORA test and debug a model?
The first model issues are identified trough the pre-processors:
- Errors in connectivity between elements.
- Errors in islands (network sections erroneously appearing as islands).
- Errors in the buses limit configuration.
- Error in the limit configuration and anomalous data in lines
- Errors in voltage level data.
- Errors in generators, loads or shunts data.
- Errors in the alarms.
- Errors in data corresponded to non-existent elements.
- Topology errors.
- Errors in the combined cycles curves.
- Errors in impedance values.
- Errors in in voltage levels assigned to non-existent substations.
Then for further model and data cleansing, the process of deploying AGORA’s State Estimator is what entails a thorough testing and debugging of an EMS model and its SCADA telemetry data.
How does AGORA deal with bad model data?
In case of having bad values for electrical parameters in any of the equipment composing the network, AGORA can be used to obtain adjustments of some parameters (as reactance of transmission lines) when many snapshots of the same model have been analyzed by the system.
Dealing with bad SCADA data
How does AGORA deal with bad measurement data?
The SCADA gives an initial indication of the quality of the measurements coming from the field, and in each estimation cycle AGORA performs a battery of electric tests to determined whether a set of measurements has good, bad or undetermined quality, in order to redefine its weights in the estimation. These electric tests are local and take place before the State Estimation properly speaking is attempted, thus ensuring the consistency of the process.
For values whose measurements are identified as of bad quality, the weights are diminished in order to have small impact in the final estimation until a satisfactory result is found.
How does AGORA deal with large areas that have no SCADA measurements?
In this case AGORA tries to estimate values in that area that will minimize the error in the area where there are SCADA measurements and where the estimation can be fully checked against the field.
In order to have realistic estimated values in the area without SCADA data, the utility must give at least limits for values in this area that will restrict the estimated values. In case no limits are available, AGORA can use pseudo-values as load profiles (or others) allowing to define limits for estimated values in that area.
It has been proven very robust, generating satisfactory estimation when SCADA driven observable measurements provide only 30-40% of the grid model employed.
When should I try to change the defaults?
How should I configure AGORA for state estimation?
How should I configure AGORA Limit Violation Solver?
How should I configure AGORA Restoration?
What reports does AGORA generate?
AGORA system configuration
When should I try to change the defaults?
We recommend to use default values when starting to work with AGORA and change the configuration of the different modules once the user know how to use them in order to better adapt the results to the company operations standards.
Does AGORA provide capability of multiple, distributed slacks?
Yes, AGORA is able to simulate correctly both primary regulation (governor droop) and secondary regulation (AGC).
State Estimator
How should I configure AGORA for state estimation?
The process of configuring the estimation can be divided into the following stages:
- Configuring the parameters which determine how the Estimator will behave: criterion for classifying measurements, activation of SE and general weighting assigned to analogue measurements.
- Assigning the Slack node: configures the bus or buses which will be considered to be balance (slack or swing) nodes during the state estimation process.
- Configuring the Advanced Estimation: configures the units for which a specific value, or their connectivity, is to be estimated (taps, shunts, connectivity, etc.)
- System configurations: these contain parameters which modify or affect how the estimation process operates. It is recommended that the default configuration is used, except for the considerations stated in the section on “tuning the Estimator parameters”.
AIA recommends that the initial adjustments are set to the default values (DEFAULT), especially as a starting point in the initial tuning. They can later be adjusted, depending on the characteristics of the network, once it has been tuned.
Limit Violation Solver
How should I configure AGORA Limit Violation Solver?
The user preference on corrective equipment choice can be configured to match the user operation standards or best practices in the Limit violation configuration menu of the general configuration menu of AGORA.
The behavior of the algorithm used in Limit Violation Solver may be personalized by the system administrator, using the following configuration screens:
- Load Weights: allows to give weightings to the units and devices in the system, by assigning a given weight to the various loads and generators.
- Parameters: the user can activate and/or assign weights to different strategies (types of action associated with types of limit violation in the network) for the limit violation solver algorithm. Costs and penalties can also be assigned to generic criteria for classifying and selecting actions.
- Atypical Cases: the user can configure a subset of important actions for which the evaluation will always be included/excluded by the limit violation solver algorithm.
Restoration
How should I configure AGORA Restoration?
The Restoration menu groups together the graphical screens used to configure the various parameters involved in the restoration process. For the system to operate, it requires a minimum configuration. This should be added when the administrator wishes to increase the degree of control over the “freedom” of the restoration algorithm.
The screens making up this menu are the following:
- Operation Area: it refers to the area controlled by the operator. It may be defined as a set of substations.
- Action execution times: during the restoration process, the execution times associated with each of the actions are very significant in the strategy that the system determines.
- Restoration Weights: the Restoration heuristic function is conditioned by seven weights. Modifying these weights changes the personality of the system and thus influences its restoration policy.
- Disturbance configuration: allows to define the load priority, rules and black start strategy to be used following the type of disturbance.
- Load Weights: in a restoration, it may be the case that not all the loads lost have the same importance, i.e. it is appropriate to return the service to certain loads, (hospitals, for example) more urgently than to other loads. This discrimination is achieved by using the restoration load weights screen.
- AGORA Plan: AGORA plans are sets of actions. In principle, they can contain any type of action required by the operator.
- Rules: define the conditions required in order to identify when the algorithm has reached the desired state, to apply these conditions and then to give the algorithm “freedom” to complete the restoration in accordance with its heuristics. The system will only “consider” the rules if all the intermediate electrical states are stable. They can be used to incorporate restraining certain actions for known dynamic problems.
- Disturbance Script: allows to configure the parameters of the script that is used to detect the condition of disturbance (minimum number of alarms and switch changes, time window to count alarms and switches, etc.).
Reports
What reports does AGORA generate?
The additional features of the AGORA system include generating reports. The purpose of these report is to export the information managed by the system.
The result of the report are ASCII files in a pre-established format, which are easy to handle using market standard tools (Excel, Awk, R, etc.).
The reports may be generated by an automatic process (depending on the activation parameters which have been configured) or manually, at the user’s request.
The report features have been structured as several configuration screens, specifically:
- Catalog: the catalogs are basic information trees. There is a catalog for each type of information where the information is structured hierarchically, in a tree graph.
- Scheduled: the system allows the generation of reports to be triggered by timing or by events.
- Packager Filters: allows to group together and tally the information in different daily reports, in order to generate monthly reports.
- Packager: based on the daily information files generated as timed reports, it is possible to group together and consolidate the information and thus generate monthly files in a predefined format for each of the types of packaging.
Interoperability with SCADA/EMS tools
How does it connect and integrate with existing SCADA/EMS?
AGORA is a system running under UNIX/Linux. It is an advanced suite of EMS applications that integrates with the existing SCADA and/or EMS, either via ICCP protocol or via ad-hoc files that are transferred under secure protocols such as ssh/scp/sftp/rsync. The users access their graphical user interface via standard X-Window connections.
AGORA is commonly deployed on three servers: one of them runs the system for simulation and training purposes, while the other two are intended for Real-Time purposes and run under a hot fail-over cluster configuration, as shown in the next figure.
Pre-processors
What are the pre-processors needed for AGORA integration with existing SCADA/EMS?
The pre-processors are specialized modules for format translations. The pre-processors translate the information between the external systems’ or standard data representations and the AGORA internal data representation.
AGORA has pre-processors for PSS/E, GE, SIEMENS EMPROS (SpectrumCC), RANGER ABB (Network Manager) and AREVA/ESCA/ALSTOM/GE (e-terraplatform) SCADA data formats and network models. It also can receive input in CIM standard formats. Specific pre-processors for other required formats can be written as necessary.
Different technologies are used in this context, such as C language modules and XLST style sheets, for example. Also, scripts for text processing engines (awk) are used for the rapid prototyping of pre-processors.
Export
Does AGORA have the capability to export estimated snapshots?
Yes, there is the capability to export estimated snapshots in different formats: GE PSLF, and SIEMENS PSSE.
Platforms and Hardware
What platforms is it available for?
The system runs under UNIX, and it is fully written in C/C++ (with some bourne shell scripts). The only third-party dependencies are the graphics libraries ILOG Views & Charts 5.x, which is the toolkit in which all graphical interfaces are written. Deployments that make use of ICCP data sources may also need third-party libraries for such type of connections.
It can run under most UNIX flavours (Linux, AIX, Solaris), although the current active ports are Linux and AIX. We currently recommend 64-bit Linux for maximum performance.
What file formats and versions are supported?
AGORA has pre-processors for PSS/E, GE, SIEMENS EMPROS (SpectrumCC), RANGER ABB (Network Manager) and AREVA/ESCA (e-terraplatform) SCADA data formats and network models. Specific pre-processors for other required formats can be written as necessary.
Do the various versions of the SCADA/EMS format get auto-detected by AGORA?
No. As a general rule, new pre-processors may be needed or the existing ones may need adaptation ad hoc. However sometimes new versions are downwards compatible in design by EMS vendors and no adaptation is needed.
What is the required and the recommended hardware to run it?
After extensive performance benchmarking, we have determined that the industry benchmarks that best predict the expected performance of AGORA are those coming from the HPC field (High Performance Computing). Namely, we use these two standard benchmark suites as a guide:
- SPEC CPU 2006 (www.spec.org): They offer two benchmarks, one for integer-dominated computing and one for floating-point computing. Each one of them has two versions, one for single-task execution and the other for throughput (SPEC_rates).
- STREAM (www.cs.virginia.edu/stream): This is an index that measures directly just the RAM bandwidth performance. It is a good complement to the SPEC numbers, for cases where the application uses much more RAM than the L2 cache size.
Experimentally we have determined that, on average, and for grid sizes of about 4,000 electrical nodes or less, AGORA’s tasks are slightly more limited by raw CPU power than RAM bandwidth, so we tend to favor the results from SPEC CPU 2006 rather than STREAMS.
Judging from current market trends, as of 2017-18, our clear recommendation is Intel Xeon architectures, with quad-core (or higher core count) microprocessors. Newer POWER8 chips from IBM could in principle be an alternative option, but for the AGORA use-case, they perform worse and cost significantly more. Other RISC competitors, such as SPARC or MIPS, are out of the question completely, since they just do not have the floating point performance needed for scientific applications such as AGORA.
Is there any preference from different vendor architectures?
At the moment we recommend systems based on Intel’s Xeon E5 series, rather than the AMD Opteron series. For about the last ten years, ever since the introduction of the Core microarchitecture, Intel has been consistently beating AMD in single-core FPU performance, which is one of the most relevant performance metrics for AGORA. However, it is also important to remark that single-core performance of modern processors has essentially peaked, for most practical purposes. This means that the only way to increase performance is through parallelism.
What should be the architecture for 10-15 simultaneous users?
For general use of an AGORA system with a maximum number of 10 to 15 simultaneous users, and grid sizes of up to 4,000 electrical nodes, we generally recommend a 2U rack-mount server with a two-processor system board (for a total of 8 cores or higher), as a sweet spot in terms of price/performance. Then this should be configured with the fastest available Intel Xeon processor, and the fastest RAM available. At the time of writing (Nov 2017), we strongly recommend systems built with the Intel Xeon E5 processor family (v4 or later), in dual-socket configuration, for maximum price/performance ratios.
What is the recommended memory and storage configuration?
The minimum recommended amount of RAM for this usage scenario is 16 GB; of course 32 GB would give more leeway in case the number of users gets closer to 15.
As for hard disk space, this is not currently a problem, as most server configurations start with more than 300 GB of available space, which is more than needed, typically1. Auxiliary external storage for keeping historical copies of information generated will be needed dependent on SEC best practices and auditing regulations. This will need a design and sizing of backup procedure during SOW (statement of work) phase.
Does AGORA have other maintenance recommendations?
Apart from performance considerations, we also recommend that the server has a “Lights Out Management” (LOM) system, in order to have remote access to the BIOS/Firmware and therefore be able to do remote diagnostics, operating system installs, etc.
Power Flow Algorithm
What is the AGORA power flow method?
It is a wholly new power flow method know as HELM™ (Holomorphic Embedding Load Flow Method), built on a radically different approach to the problem. It uses advanced concepts from Complex Analysis (such as algebraic curves and Padé approximants). In practical terms, the benefit is that it is a reliable and unattended method: without having to provide an initial point, it finds the desirable operating solution of the system when the case is feasible; and conversely, it unambiguously yields “no solution” when the case is infeasible (voltage collapse). In other words, its results are always unequivocal. It does away with the inherent problems of convergence and fractality in iterative algorithms.
The traditional method for solving power flow problems involves a second-order iterative algorithm (Newton-Raphson), sometimes using an approximate updating of the Jacobian matrix (such as in the Fast Decoupled Load Flow, FDLF) to speed-up the calculation. The NR and FDLF methods are efficient, if the initial staring point is sufficiently close to the correct solution. However, if the initial point is “too far” from the current solution, convergence cannot be assured. In many cases, expert human intervention is required to effectively solve this problem.
AGORA’s approach is new and different than the Newton-Raphson method and variations (FDLF, Continuation Power Flow, etc.). Its methodology:
- Is non-iterative
- Does not require an initial point
- Finds the solution even when the network condition is close to voltage collapse
AGORA’s methodology is based on the HELM™ method, Holomorphic Embedding Load flow method, that has been granted three USA Patents[2]. More info can be found at:
http://en.wikipedia.org/wiki/Holomorphic_embedding_load_flow_method
This method guarantees accurate and correct solutions: HELM’s solution is always the most desirable operating point of the power system (i.e. it will never yield so-called low-voltage solutions). It is a method derived from concepts in theoretical physics, is robust, and its performance is competitive with NR and FDLF methods.
[2] US Patents Nos. 7519506, 7979239 and 8849614.
Where is HELM™ method used in AGORA?
It is used as a kernel function almost everywhere:
- Used as a last verification in the State Estimation to guarantee solving the load flow equations.
- Used in the simulation on-line for an operator power flow for “what if” questions and future supervised actions effects.
- Used extensively in the Contingency Analysis and PV-QV curves functionalities.
- Thousands of load flows are computed in the exploration of paths in the Limit Violation Solver and Restoration action plans generating procedures.
Algorithmic methodologies
What module uses A* algorithm?
The Restoration module is the main function relying on the A* algorithm. The A* (pronounced “A-star”) algorithm generates the sequence of manoeuvres that will bring the network to the state close to the pre-disturbance state, taking into account the system equipment that is unavailable.
Other modules such as Limit Violation Solver also employ it.
Could you briefly explain the A* algorithm, which is central to AGORA’s Restoration?
It belongs to the family of optimal path algorithms in a directed graph. In computer science, A* (pronounced “A star”) is a best-first, tree search algorithm that finds the least-cost path from a given initial node to one goal node (out of one or more possible goals) [3].
It uses a distance-plus-cost heuristic function (usually denoted f(x)) to determine the order in which the search visits nodes in the tree. The distance-plus-cost heuristic is a sum of two functions: the path-cost function (usually denoted g(x), which may or may not be a heuristic) and an admissible “heuristic estimate” of the distance to the goal (usually denoted h(x)). The path-cost function g(x) is the cost from the starting node to the current node.
Since the h(x) part of the f(x) function must be an admissible heuristic, it must underestimate the distance to the goal. Thus for an application like routing, h(x) might represent the straight-line distance to the goal, since that is physically the smallest possible distance between any two points (or nodes for that matter).
The algorithm was first described in 1968 by Peter Hart, Nils Nilsson, and Bertram Raphael. In their paper, it was called algorithm A. Since using this algorithm yields optimal behavior for a given heuristic, it has been called A*.
[3] Source: Wikipedia.
What is the objective of the A* algorithm?
The purpose of the A Star (A*) task is to provide the sequence of actions that will restore the state of the network to the pre-disturbance state (to the extent allowed by the unavailable equipment, if any).
The first step is to determine the changes produced in the network between the pre-disturbance (or “normal”) state and the disturbance (current state) in order to obtain the list of actions or manoeuvres that must be accomplished to restore the system. The main difficulty of this task is not to obtain the list of actions, but to establish the order in which they should be executed. It usually isn’t viable to explore all possible permutations of actions to restore the system in a reasonable amount of time, even for small disturbances. For this reason, the task uses the A* technique in order to efficiently calculate the restoration plan.
A* calculates a final pseudo-state that gives orientation towards which the restoration plan should be directed.
How does AGORA adapt the A* algorithm?
The A* algorithm is used for the search of optimal paths in a space of states, where a departure state and one target state is identified. In AGORA, the departure state identified as the disturbance state and the objective or target electric state is the pre-disturbance state (or a state very close to it). AGORA does this by defining many sub-objective states between departure and target states and applying A* algorithm between each sub-objective states. The algorithm ends when the target state is reached by the constructed path.
Passing from one state to another is done through a transition (actions or manoeuvres in the equipment). In this case a transition implies an action or an indivisible set of action in the electrical network. The application of action on the electrical state before the transition results in the post-transition electrical state.
The main difficulty of the problem is the definition of the heuristic giving weights of all possible paths between two electric states according to the user criteria coherent with the utility best operational practices and compliant with regulator norms as well, and establish the order in which the necessary actions should be executed (which is done using again A* technique in order to efficiently calculate the restoration plan).
As in any other AI technique, finding the right heuristics first and then refining them with knowledge of best operational practices is a team work with expert craftsmanship. This work has been achieved along many years of team work with clients. Default heuristic today works well in all grids, and minor customization of months not years need to take place.
Does the Restoration module use algorithms other than A*?
In addition to the A*, other algorithms are used. Mainly those including: min-cost network flow, to calculate heuristics; dynamic programming, to calculate heuristics and network unmeshing; an algorithm to correct limit violations during the restoration process; and load flow or load distribution to obtain voltages and flows for the evaluation of the different restoration paths.
Performance methodologies
Can AGORA improve performance by parallel computing?
Fortunately AGORA has always been built to take advantage of parallelism in a very natural way, because all user sessions are implemented as independent UNIX processes. Only the State Estimator and a few house-keeping data-management processes are global in scope. The Restoration module is the only CPU-heavy algorithm that is also global in scope; however, it is threaded and therefore it can take advantage of modern multi-core architectures. Therefore, when choosing hardware for AGORA, a good rule of thumb is to always specify the fastest available CPU, and then vary the total number of CPU cores according to the expected number of simultaneous users and scheduled background tasks (contingency analyses, etc.). Of course, hardware scalability has its limits, which these days originate mainly from concurrent accesses to RAM. Our technical spec notes provide advice in choosing the right hardware and the right number of CPUs and RAM for a given set of user requirements.