The goal of this software is to automatically generate C/C++ code which reads and writes GOOSE and Sampled Value packets. Any valid IEC 61850 Substation Configuration Description (SCD) file, describing GOOSE and/or SV communications, can be used as the input. The output code is lightweight and platform-independent, so it can run on a variety of devices, including low-cost microcontrollers. It's ideal for rapid-prototyping new power system protection, control, and automation systems that require communications.

This readme file describes how to set up the software, and its basic use.

The code is meant to be a proof of concept, and is highly experimental. It has not been tested on many SCD files. The code is also still in development at the moment, so some features may be broken or incomplete.

• Implements sending and receiving GOOSE and Sampled Value packets
• Lightweight, and suitable for low-cost microcontrollers
• Platform-independent, and any C/C++ compiler should work
• Performs validation of the SCD file, and reports any problems
• Can optionally support fixed-length GOOSE encoding, which reduces GOOSE encoding time by approximately 25-50%
• Supports initialisation of data type values, and instance-specific values
• The platform can be used in two ways:
  • As part of a native C/C++ program. This approach would be used where deterministic real-time performance is important, or where the the network interface is custom (such as on a microcontroller). It also works well with the Qt C++ GUI framework.
  • As part of a Python or Java program. This approach uses additional C code (with winpcap/libpcap) to handle the communications and data model, with SWIG wrappers to link to a Python or Java program. It is useful for any application where sub-millisecond performance is not needed, because it offers the comfort and convenience of writing your control logic code in a high-level language.
• Open source, under the GPL 2

This process has been tested on Windows and Ubuntu, but other Linux flavours and OS X should work too.

The software requires Eclipse, with the Eclipse Modeling Framework (EMF). Java Emitter Templates (JET) is needed for development, but not to run the code. It's easiest to start with the version of Eclipse that comes with the Modeling Tools bundle (see here: http://www.eclipse.org/downloads/). (If you are planning on using the Python or Java interfaces on Windows, it is best to use the 32-bit versions of Eclipse, and the JDK.)

There are two source code trees: emf (in Java), and c (obviously written in C). Each should be a separate project in Eclipse. The Java emf project directory is organised as follows:

• src/
  • scdCodeGenerator/: code that does the bulk of the conversion from an SCD file to C code. The class Main contains the main() function for the project, and contains the filename for the input SCD file.
  • scdCodeGeneratorTemplates/: template classes that are generated by JET.
  • ch/: the EMF Java model implementation. These files are all automatically generated by EMF, but are included in the repo for convenience.
• model/: the IEC 61850 XML Schema files. EMF uses these to generate the model.
• templates/: the template source files used by JET.

1. Start Eclipse, with the Workspace set to the root of the repository directory, e.g., /home/user/rapid61850 on Linux.
2. Create an "EMF Project" called "emf", at the location of the repository code.
3. Select "XML Schema" as the Model Importer type. Select all the IEC 61850 XML Schema documents in the emf/model directory.
4. Select the three root packages that are imported (although, only scl is used). Click "Finish". This will re-generate some files in emf/model: scl.ecore, lcoordinates.ecore, lmaintenance.ecore, and SCL.genmodel.
5. Create a new project of type "Convert Projects to JET Projects", and select the emf project. For the emf project, go to Project Properties > JET Settings, and set Template Containers to "templates", and Source Container to "src". Delete the scdCodeGeneratorTemplates directory in the root of emf that was created before JET was configured correctly.
6. Open SCL.genmodel and right-click on the root of the model tree. Select "Show Properties View" and ensure that "Compliance Level" is set to "6.0". Right-click on the root again and select "Generate Model Code". This should re-generate the model implementation files (in the emf/src/ch directory), and set up the project properly for using the generated code.
7. Two additional JAR libraries must be included for the project to compile. In the Project Properties for emf, go to Java Build Path > Libraries. Click on "Add External JARs..." and find com.ibm.icu_4.4.2.v20110823.jar and org.eclipse.emf.query_1.2.100.v200903190031.jar (or similar versions). These should be located in the "plugins" directory within the Eclipse installation.

An example SCD file and a main.c file are provided. Many of the other C files are generated automatically. For the C code to compile with Eclipse, you should:

• Install MinGW and add C:\MinGW\bin; to PATH in the Project Properties > C/C++ Build > Environment options. (Other compilers should work too.)
• In Project Properties > C/C++ Build > Settings > GCC Compiler Includes, set "${workspace_loc:/${ProjName}/Include}" as an include path.
• In Project Properties > C/C++ Build > Settings > MinGW C Linker, add wpcap and ws2_32 (assuming you are using Windows) to "Libraries" and add "${workspace_loc:/${ProjName}/Lib}" and "C:\MinGW\lib" to "Library search path".
  • With Linux, use pcap instead of wpcap, and just add "${workspace_loc:/${ProjName}/Lib}" to the "Library search path".
• The WinPcap library files and header files (from http://www.winpcap.org/devel.htm) have been included in the repository for convenience. The PC must also have the WinPcap driver installed (either by installing Wireshark, or from http://www.winpcap.org/install/default.htm).
  • With Ubuntu, libpcap can be installed using sudo apt-get install libpcap-dev.
  • Remember that, on Linux, libpcap needs to run as root, so either start Eclipse or run the compiled binary from the Terminal with sudo. Alternatively, you can grant the binary the capability to access the network interface using: sudo setcap cap_net_raw,cap_net_admin=eip /path_to_project/rapid61850/c/Release/c.

First, open the file Main.java. In the Main class, set the value of SCD_FILENAME to the filename of the SCD file. The SCD file should be in the same directory as the Main.java file. Run the Java project to generate the C implementation. If the SCD parser complains, ensure that the first two lines of the SCD file exactly match those from the example scd.xml in the repository. It's usually best to refresh the C project in Eclipse, to ensure that Eclipse knows about the new or modified files.

A basic C main() function will look something like:

#include "iec61850.h"

int length = 0;
unsigned char buffer[2048] = {0};

int main() {
	initialise_iec61850();											// initialise all data structures

	// send GOOSE packet
	E1Q1SB1.S1.C1.TVTRa_1.Vol.instMag.f = 1.024;					// set a value that appears in the dataset used by the "ItlPositions" GOOSE Control
	length = E1Q1SB1.S1.C1.LN0.ItlPositions.send(buffer, 1, 512);	// generate a goose packet, and store the bytes in "buffer"
	send_ethernet_packet(buffer, length);							// platform-specific call to send an Ethernet packet


	// in another IED...


	// receive GOOSE or SV packet
	length = recv_ethernet_packet(buffer);							// platform-specific call to receive an Ethernet packet
	gse_sv_packet_filter(buffer, length);								// deals with any GOOSE or SV dataset that is able to be processed

	// read value that was updated by the packet (it will equal 1.024)
	float inputValue = D1Q1SB4.S1.C1.RSYNa_1.gse_inputs_ItlPositions.E1Q1SB1_C1_Positions.C1_TVTR_1_Vol_instMag.f;

	return 0;
}

The data structures used for generating GOOSE and SV packets are stored within LN0. GOOSE packets are generated by calling the appropriate send(buffer, statusChange, timeAllowedToLive) function, where statusChange should be 1 if any value in the dataset has changed, and where timeAllowedToLive is the time in milliseconds for the receiver to wait for the next re-transmission. SV packets are sent by calling the update(buffer) function, which returns 0 if the next ASDU was written, but other ASDUs are free. It returns the size of the packet when all ASDUs have been written (and buffer contains the packet data). Clearly, a real implementation might include the use of platform-specific timers, interrupts and callbacks, where needed.

The generated C code implements all IEDs specified in the SCD file. You can use the code to emulate the communications between several IEDs, or just use one IED's implementation.

Callbacks should be set up in the form:

void SVcallbackFunction(CTYPE_INT16U smpCnt) {
	;
}

void GSEcallbackFunction(CTYPE_INT32U timeAllowedToLive, CTYPE_TIMESTAMP T, CTYPE_INT32U stNum, CTYPE_INT32U sqNum) {
	;
}

//...

D1Q1SB4.S1.C1.exampleMMXU_1.sv_inputs_rmxuCB.datasetDecodeDone = &SVcallbackFunction;
D1Q1SB4.S1.C1.RSYNa_1.gse_inputs_ItlPositions.datasetDecodeDone = &GSEcallbackFunction;

where D1Q1SB4.S1.C1.exampleMMXU_1 is a Logical Node defined in datatypes.h (and ied.h). rmxuCB is the name of the SampledValueControl, in a different IED, which sent the SV packets. After being initialised, the callback function will be executed after this dataset is successfully decoded, to allow the LN to deal with the new data. For example, by default, only one packet of data is saved for each GSE or SV Control - and is overwritten when a new packet arrives. Therefore, it may be useful to use the callback to log the data to a separate memory buffer.

To enable fixed-length GOOSE encoding, in ctypes.h set the value of GOOSE_FIXED_SIZE to 1. Otherwise, it should have a value of 0. This can only be enabled globally for all GOOSE encoding, rather than on a per Control basis.

All platform-specific options can be edited in ctypes.h or ctypes.c. For example, for a big endian platform, change:

#define LITTLE_ENDIAN		1

to:

#define LITTLE_ENDIAN		0

All CTYPE_* definitions must map to local datatypes of the correct size and sign.

In ctypes.c, the basic library function memcpy() is used to copy bytes in order (according to platform endianness), and reversememcpy() copies the bytes of multi-byte data types in reverse order (for converting between endianness). Although these should work, they can be replaced with platform-specific alternatives for better performance.

The value of TIMESTAMP_SUPPORTED should be set to 0, unless generating timestamps has been implemented for your platform. An implementation for Windows has been included by default.

So far, this readme has described how to use the native C/C++ interface. It's also possible to use SWIG to automatically generate wrappers for high-level languages from C/C++ header files. At the moment, Python and Java interfaces on Windows and Linux have been tested, but other languages (such as C#, Lua, Perl, Ruby, etc.) should work too.

Four C files, with filenames interface*, are generated along with the rest of the GOOSE/SV code. These files, and the SWIG interface file rapid61850.i, are used as the input to SWIG. They contain functions to start a (platform-dependent) network interface using winpcap/libpcap, and functions to send GOOSE or SV packets using that network interface. All of the interaction with pcap is done in C, and is hidden by the interface given to SWIG.

Note that this interface can also be used within a C/C++ application - this is shown in the example main.c file, if HIGH_LEVEL_INTERFACE is defined as 1. If your are not using this high-level interface, and are using the plain C interface, you may need to exclude the two interface*.c files from the build in Eclipse.

If using MinGW as the C compiler (as described above), this process is significantly simpler if the 32-bit versions of Eclipse and the JDK are used. The following instructions assume this. It's also assumed that the Python or Java application exists within a directory at the same level as the emf and c directories.

• Generate the C code for your SCD file, as described above.

• Download swigwin, which is a pre-compiled binary of SWIG for Windows. Once unzipped, there are two options for using this:

  • Add the location of swig.exe to the Windows PATH environment variable.
  • Or, copy the contents of the swigwin directory (i.e., copy swig.exe and all the sub-folders) to the c/src directory. You will need to tell Eclipse to exclude these directories from the build.

• Create the directory for your Python or Java program called, for example, python_interface or java_interface. You may wish to make this an Eclipse PyDev or Java project.

• Open a command prompt at the c/src directory, and run SWIG using one of the following commands:

  For Python:

   swig -python -outdir ..\..\python_interface rapid61850.i
  

  For Java:

   swig -java -outdir ..\..\java_interface rapid61850.i
  

Now we need to change the compiler settings for the c project to generate a dynamic library, instead of an executable. This differs for Python and Java. It may be helpful to create different build configurations in Eclipse if you need to use more than one of the C/C++, Python, or Java interfaces. You may also need to exclude the existing main.c file from any Python or Java builds.

• In C/C++ Build > Settings > Build Artifact:

  • set Artifact Type to Shared Library
  • set Artifact name to rapid61850
  • set Artifact extension to pyd
  • set Output prefix to _

• In C/C++ Build > Settings > Tool Settings > Includes, use the following Include Paths (adjust these to match the exact version and location of Python on your system):

  • "C:\Python27"
  • "C:\Python27\include"
  • "C:\Python27\Lib"
  • "${workspace_loc:/${ProjName}/Include}"

• In C/C++ Build > Settings > Tool Settings > Libraries, use the following Libraries (-l):

  • wpcap, python27, and ws2_32. (Again, adjust python27 to the correct version.)

• In C/C++ Build > Settings > Tool Settings > Libraries, use the following Libraries search paths (-L):

  • "${workspace_loc:/${ProjName}/Lib}"
  • "C:\Python27\libs"

• Build the C project, and copy the _rapid61850.pyd file from the Release folder to the python_interface project directory.

• Create and run your Python code, e.g.:

  import rapid61850
  from rapid61850 import *
  
  rapid61850.start()
  
  rapid61850.gse_send_D1Q1SB4_C1_MMXUResult_buf(1, 512)     # send GOOSE packet
  
  rapid61850.cvar.E1Q1SB1.S1.C1.LPHDa_1.Mod.stVal = MOD_ON  # interact with IED data model
  print rapid61850.cvar.E1Q1SB1.S1.C1.LPHDa_1.Mod.stVal

  Note that all C global variables appear within rapid61850.cvar.

• In C/C++ Build > Settings > Build Artifact:

  • set Artifact Type to Shared Library
  • set Artifact name to rapid61850
  • set Artifact extension to dll
  • leave Output prefix blank

• In C/C++ Build > Settings > Tool Settings > Includes, use the following Include Paths (adjust these to match the exact version and location of Java on your system):

  • "C:\Program Files (x86)\Java\jdk1.7.0_03\include"
  • "C:\Program Files (x86)\Java\jdk1.7.0_03\include\win32"
  • "${workspace_loc:/${ProjName}/Include}"

• In C/C++ Build > Settings > Tool Settings > Libraries, use the following Library search path (-L):

  • "${workspace_loc:/${ProjName}/Lib}"

• In In C/C++ Build > Settings > Tool Settings > Miscellaneous, add -Wl,-add-stdcall-alias to the Linker flags

• Build the C project, and copy the rapid61850.dll file from the Release folder to the java_interface project directory.

• Create your Java code, e.g.:

  public class Main {
      static {
          System.loadLibrary("rapid61850");
      }
  
      public static void main(String[] args) {
          rapid61850.start();
  
          System.out.println(rapid61850.gse_send_E1Q1SB1_C1_Performance_buf(1, 512));                     // send GOOSE packet
  
          rapid61850.getE1Q1SB1().getS1().getC1().getMMXUa_1().getMod().setStVal(Mod.MOD_ON);             // interact with IED data model
          System.out.println(rapid61850.getE1Q1SB1().getS1().getC1().getMMXUa_1().getMod().getStVal());
      }
  }

• To run the Java program, you first need to specify the path to the native library. In Project Properties > Java Build Path > Libraries, expand the "JRE System Library" tree and select "Native library location". Click on "Edit..." and enter the project name (e.g., java_interface) as the Location path. Note that if the interface changes, such as due to changes in the SCD file, then all .java and .class files generated by SWIG should be deleted before a new dynamic library is compiled and used by the Java program.

On Linux, it's easier to create the Python or Java interface with the Terminal, rather than with Eclipse. The steps below assume Ubuntu (and have only been tested on 11.10 64-bit), so it may differ on other distributions.

Install the following packages:

sudo apt-get install libpcap-dev
sudo apt-get install swig
sudo apt-get install build-essential
sudo apt-get install python2.7
sudo apt-get install openjdk-6-jdk

Open a Terminal at the rapid61850/c/src directory.

# attempt to clean up any previous files
rm *.o *.so *_wrap.c rapid61850.py rapid61850.pyc

# run SWIG, output goes in current directory
swig -python rapid61850.i

# compile and link the C library
gcc -fPIC -c *.c -I/usr/include/python2.7
gcc -shared *.o -lpcap -o _rapid61850.so

# run Python. sudo is needed for the network interface
sudo python2.7

# example Python program:
>>> import rapid61850
>>> rapid61850.start()
>>> print rapid61850.gse_send_D1Q1SB4_C1_MMXUResult_buf(1, 512)
332
>>> exit()

# attempt to clean up any previous files
rm *.o *.so *_wrap.c java/*.class java/*.java

mkdir java    # only needed once

# run SWIG, and put the .java files (there will be a lot) in the "java" sub-directory
swig -java -outdir java rapid61850.i

# compile and link the C library
gcc -fPIC -c *.c -I/usr/lib/jvm/java-6-openjdk/include -I/usr/lib/jvm/java-6-openjdk/include/linux
ld -G *.o -lpcap -o librapid61850.so

# compile all .java files, including the sample program
javac -d java/ java/*.java ../../java_interface/Main.java

# run the Java program. sudo is needed for the network interface
cd java
sudo java -Djava.library.path=/home/steven/rapid61850/c/src/ Main    # this path must be set correctly

• Some data types are not supported yet. However, the main useful data types (integer, floating-point, and boolean) are supported.
• FCDAs and ExtRefs cannot use the syntax "vector.mag.f" as values for data object or data attribute references.
• Data types cannot contain arrays.
• According to the standard, SV datasets should only contain primitive data types, and not constructed types. However, because SV encoding involves fixed-length value fields, it is always possible to reconstruct the data, if encoded and decoded consistently. Therefore, this library will allow constructed types to be encoded in SV packets. Semantically, SV datasets should only contain data values that have been sampled at the specified sampling rate. Again, for practicality, this library allows any DA or DO to be used in SV datasets.